Research ArticlePalmprint Authentication System Based on Local and GlobalFeature Fusion Using DOST

N B Mahesh Kumar and K Premalatha

Bannari Amman Institute of Technology Sathyamangalam Erode Tamilnadu 638 401 India

Correspondence should be addressed to N B Mahesh Kumar mknbmaheshkumargmailcom

Received 23 September 2014 Revised 16 December 2014 Accepted 16 December 2014 Published 31 December 2014

Academic Editor Qiankun Song

Copyright copy 2014 N B M Kumar and K Premalatha This is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

Palmprint is the region between wrist and fingers In this paper a palmprint personal identification system is proposed based onthe local and global information fusionThe local and global information is critical for the image observation based on the results ofthe relationship between physical stimuli and perceptionsThe local features of the enhanced palmprint are extracted using discreteorthonormal Stockwell transformThe global feature is obtained by reducing the scale of discrete orthonormal Stockwell transformto infinity The local and global matching distances of the two palmprint images are fused to get the final matching distance of theproposed schemeThe results show that the fusion of local and global features outperforms the existing works on the available threedatasets

1 Introduction

The unique physical or behavioural characteristics of humanbeings are of broad interest in the biometrics based methods[1] In the modern e-world biometrics has great potentialsbecause of their high accuracy and convenience to useResearchers and scientists have systematically investigatedthe use of a number of biometric characteristics like fin-gerprint [2ndash5] face [6] iris [7 8] palmprint [9 10] handgeometry [11] hand vein [12] finger knuckle-print [13 14]voice [15] and ear [16] in the development of computingtechniques The use of applications like automatic personalauthentication systems in day-today life results in reliable andeffective security control Recently hand-based biometricshas been attracting significant attention in the modern e-world

The region between wrist and fingers is referred to aspalmprint It has features like texture wrinkles principlelines ridges and minutiae points that can be used for itsrepresentation The critical properties of biometric char-acteristics such as universality individuality stability andcollectability are satisfied by palmprint compared to otherbiometric traits It has relatively stable and unique features Itis nonintrusive and data collection is very easy It needs fewer

cooperations to collect data from persons It provides highefficiency using low resolution images taken from the lowcost device Recently palmprint based recognition systemshave received a wide attention from researchers and effortshave been made to build a palmprint based system basedon structural features of palmprints such as principle lineswrinkles datum points minutiae points ridges and creasepoints

Features extracted from the palmprint image can beclassified into local and global features Local features char-acterize the main topological characteristics of a small partof the trajectory Global features characterize the relationshipof different line segments within a trajectory The overallcontent and shape of the objects in the image represent theglobal features The specific information in the local regionsis defined as local featuresThe advantages of local features aredefined in terms of template size and its discriminability Butthe local features also have inevitable drawbacks in practicalusage Due to the sensitivity of the palmprint image it isdifficult to exactly obtain the local features The performanceof the local features degrades due to its poor palmprint qualityand capture area The global features represent the palmprintin a global perspective Most of the global features arecontinuous and smooth everywhere expect in some special

Hindawi Publishing CorporationJournal of Applied MathematicsVolume 2014 Article ID 918376 11 pageshttpdxdoiorg1011552014918376

2 Journal of Applied Mathematics

Image acquisition Preprocessing Local and global feature extraction

Enrollverify

identify

Decision Matching

Database

Figure 1 Block diagram of the proposed palmprint based recognition system

regions The global features can be extracted more consis-tently from the poor quality of the partial palmprints It istoo space and time consuming to directly store and comparethe features pixel by pixel Global feature based matchingoverlaps two given templates with different transformationparameters and estimates the similarity score between thecorresponding cells Compared to local features the globalfeatures have less distinctness and hence been often exploitedtogether with other features or in the preprocessing stageof palmprint matching Both the local and global featuresare fairly independent and capture current information andhence it is realistic to improve the discriminating ability ofmatching by fusing features The local and global features arecombined in the matching stage with accessible feature levelfusing strategies The combined local and global features canimprove the performance of the palmprint systems on largescale databases How to select features is pivotal for the effectof feature combinationThe combination of irrelative featureswill improve the accuracy or efficiency of the palmprintbiometric systems The suitable hierarchical approach canbe employed to reduce the additional time or memory costrequired while fusing the local and global features

In this work a local-global feature fusion technique forpalmprint biometric system is proposed The instantaneousphase information is extracted as the local feature by DOSTThe global feature can be obtained by aggregating the scaleof the DOST The DOST can be reduced to the Fouriertransform of the whole image if the scale of the DOST isincreased to infinity Hence the local features cannot beobtained but the finest resolution for the global frequencyanalysis of the image can be obtained Therefore Fouriertransform coefficients are obviously taken as the globalfeaturesThe arrangement between intraclass palmprint ROIscan also be refined by the global Fourier features The twomatching distances are calculated by matching the local andthe global information separately Then the two matchingdistances are merged by fusion rule to get the final matching

distanceThe block diagram of the proposed system is shownin Figure 1

The rest of this paper is organized as follows Section 2describes the review on palmprint authentication Section 3describes the overview of discrete orthonormal Stockwelltransform (DOST) Preprocessing and ROI extraction isdiscussed in Section 4 Section 5 illustrates the local andglobal feature extraction and matching Section 6 presentslocal-global information fusion for palmprint recognitionExperimental results and discussion are demonstrated inSection 7 Section 8 reports the conclusion

2 Literature Review

In [4] a new BioHashing technique is proposed to generatea cancellable template in the context of unordered set ofnoisyminutiae features and shows that the work is efficient inrealistic context In [17] an intersession variability modelling(ISV) and joint factor analysis (JFA) for face authenticationare proposed to achieve a least significant error rate reductionfor measuring accuracy The line features in palmprint areextracted by using sobel and morphological methods [10]In [18] a new chaff point generation technique for the fuzzyvault in biocryptosystems is proposed and shows that theproposed technique achieves more chaff points compared toKhalil-Hanirsquos algorithm In [19] the principle lines alongwithisolated points are used as features The ridges containingcreases of the palmprint are eliminated and it has beenproposed in [20] The features like end points of principlelines have been used in [9] The end points in the palmprintare found to be directional invariant The crease points thatare related to diseases of a person have been suggested inthe palmprint biometrics system The directional line energyfeatures are also used to identify the palmprint of a personFurther the palmprint biometric systems that are basedon structural features are not invariant to occlusion of thepalmprint image

Journal of Applied Mathematics 3

The feature extraction and patternmatching of palmprintimages are focused mainly The features in the palmprintimages can be classified as local and global features basedon the range of pixels involved in the feature extraction Thelocal feature is a quantity calculatedwithin the local patch andscrambling the detailed behaviours within this specific areaGlobal information is obtained from all or most of the pixelsin the palmprint image and the universal characteristics of theobserved image can be examined Hence palmprint biomet-rics systems can be classified into local based approaches andglobal based approaches according to these definitions Thelocal-global combination for palmprint recognition has beendiscussed only in few papers The local and global featuresare critical for the image observation and identification ofpersons as per the theories in the field of psychophysicsand neurophysiology [22] A global feature reproduces theuniversal features of the image and is appropriate for roughrepresentation A local feature scrambles the more completeinformation within the specific region Therefore local fea-ture is suitable for better representation Hence the improvedrecognition accuracy is obtained if the local and global featurecan be appropriately combined It is already used in biomet-rics systems like iris face and fingerprint The conventionalDaugman-like classifier uses global features enclosed by zero-crossing boundaries in the palmprint biometrics systemsIn two-level palmprint matching scheme [23] the globalfeatures for coarse level filtering can be extracted by Houghtransform and for fine-level matching the local featuresextracted from the localities and directions of the principallines in the palmprint

The local and global features are obtained from thenonnegative matrix factorization technique and principlecomponent analysis in [24] can be combined for palmprintbiometric system In face recognition [6] the global andlocal information are extracted by principle componentanalysis (PCA) and Haar wavelets Global features in the faceimages are extracted by keeping the low-frequency Fouriercoefficients and the local information is extracted usingGabor filters The local and global information combinationwas also utilized in the fingerprint biometrics system [25]

3 Overview of Discrete OrthonormalStockwell Transform

The S-transform is more powerful than other multiresolutiontechniques like short time Fourier transform (STFT) andwavelet transform (WT) The phase of the S-transform refer-enced to the time origin provides useful and supplementaryinformation about spectra that is not available from locallyreferenced phase information in the continuous wavelettransform (CWT)

The S-transform is advantageous for the analysis ofpalmprint images as it preserves the phase information usinglinear frequency scaling

(1) However the major limitation of S-transform is itshigh time and space complexity due to its redundantnature which makes it impractical in many cases

(2) The 2D-ST of an array of size119873 times 119873 has a computa-tional complexity of O (119873

4+ 1198734 log119873) and storage

requirements of O (1198734)(3) To eliminate this problem of 2D-ST we use DOST

which is also a multiresolution technique for extrac-tion of features from the palmprint images and isbased on the S-transform

(4) DOST have less computational and storage complex-ity in comparison to the S-transform because it usesan orthonormal set of basis functions while retainingall the advantageous properties of S-transform

(5) 2D-DOSTprovides a spatial frequency representationof an image with computational and storage com-plexity asO (1198732+1198732 log119873) andO (1198732) respectively

With the dyadic sampling scheme in order 0 1 2 log2119873minus

1 DOST of an119873 times119873 palmprint image 119891(119909 119910) is performedby the following steps

(1) Two-dimensional Fourier transform (FT) is appliedto the image119891(119909 119910) to obtain Fourier samples119865(119906 V)

(2) Partition the Fourier sample 119865(119906 V) and multiply itby the square root of the number of points in thepartition and perform an inverse FT Then the voiceimage is calculated as

119878 (1199091015840 1199101015840 V119909 V119910) =

1

radic2119875119909119875119910minus2

2119875

119909minus2minus1

sum

119906=minus2119901119909minus2

2119875

119910minus2

minus1

sum

V=minus2119875119910minus2119865 (119906 + V

119909 V + V

119910)

here V119909= 2119901119909minus1+2119875119909minus2

V119910= 2119901119910minus1+2119875119910minus2

1198902120587119894

(1199061199091015840

2119875119909minus1+

V1199101015840

2119875119910minus1)

(1)

where V119909and V119910are the horizontal and vertical voice

frequencies where 1199091015840 corresponds to 1199091015840-coordinate

in space where 1199101015840 corresponds to 1199101015840-coordinate in

space(3) Thus the features of the palmprint images are

obtained after the transformation

4 Preprocessing and ROI Extraction

In this section the acquired hand image is preprocessed andthe palmprint region is extracted The square area insidethe palm region of the hand image is considered as thepalmprint or region of interest (ROI) Normally the extractedpalm region has the rotation and translation error In orderto minimize the rotation and translation error the ROIalgorithm identifies gaps between fingers The gaps betweenthe fingers are used as the reference points to align theimage The images of the extracted palm region are alignedrotationally by levelling the gap between the index finger

4 Journal of Applied Mathematics

(a) (b)

Figure 2 (a) Sample input image (b) ROI image for PolyU palmprint database

(a) (b)

Figure 3 (a) Sample input image (b) ROI image for COEP palmprint database

(a) (b)

Figure 4 (a) Sample input image (b) ROI image for IIT Delhi palmprint database

(IF) and the middle finger (MF) and that between the MFand the ring finger (RF) Among the three gaps the gapbetween MF and RF is the toughest point In order toeliminate the translation error the gap between MF and RFis used as the reference point The maximum square regionextracted from the palm area is selected as the ROI Thesquare region is horizontally centered on the axis runningthrough the gap between MF and RF Due to the regular andcontrolled uniform illumination conditions during imagecapture the acquired hand image and its background contrastin colour The sample input image is shown in Figures 2(a)

3(a) and 4(a) for PolyU palmprint database COEP palm-print database and IIT Delhi palmprint database Globalthresholding has been applied to extract the hand from thebackgroundThe image is processedwith opening and closingmorphological operations to remove any isolated small blobsor holes Contour-tracing algorithm is applied to extract thecontour of the hand image from the palmprint The ROIextraction technique of IIT Delhi palmprint database PolyUpalmprint database and COEP palmprint database are sameas in [9 26] The corresponding Figures 2(b) 3(b) and 4(b)are shown in this paper which gives the details of ROI image

Journal of Applied Mathematics 5

5 Local-Global FeatureExtraction and Matching

51 Local Feature The feature extraction technique in anypalmprint biometric system is used to obtain good interclassseparation in minimum time The features of palmprintare extracted by using 2D-DOST after the completion ofpreprocessing and ROI segmentation from palmprint Thelocal variation of instantaneous phase is used to extractfeatures from palmprint Instantaneous phase obtained usingDOST is the resolution of phase with respective to time Ithas the merits of great accurateness sturdiness to brightnessvariation and fast matching When matching two palmprintimages the angular distance based on the normalized Ham-ming distance is used

52 Global Feature DOST is used to extract the instanta-neous phase information and also it can be considered as awindowed Fourier transform The DOST can be consideredas a windowed Fourier transformThe phase-only correlation(POC) and band-limited phase-only correlation (BLPOC)are utilized to match the global information of the twopalmprint images

521 Phase-Only Correlation (POC) The Fourier transformcoefficients are considered as the global information Thetwo given Fourier transforms are matched using POC POCbased techniques have been widely used in image registrationtasks Recently a POC technique considered as a resemblancemeasure in certain biometrics systems [7 22] BLPOC ismore effective when compared to the conventional POCBLPOC is utilized to estimate the translation parametersbetween palmprint ROIs and also measure the resemblanceof the Fourier transforms of the aligned ROIs POC is anoperational method to estimate the displacement parametersbetween two palmprint images in the Fourier transformdomain Its primary principle is the translation property ofthe Fourier transforms

Consider two 1198731times 1198732images 119891(119899

1 1198992) and 119892(119899

1 1198992)

where the assumption is that the index ranges are 1198991

=

minus1198721 119872

1(1198721gt 0) and 119899

2= minus119872

2 119872

2(1198722gt 0)

For mathematical simplicity 1198731

= 21198721+ 1 and 119873

2=

21198722+ 1 Let 119865(119896

1 1198962) and 119866(119896

1 1198962) denote the 2D discrete

Fourier transforms (2D DFTs) of the two images 119865(1198961 1198962)

and 119866(1198961 1198962) are given by

119865 (1198961 1198962) = sum

11989911198992

119891 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119865(11989611198962)119890119895120579119865(11989611198962)

119866 (1198961 1198962) = sum

11989911198992

119892 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119866(11989611198962)119890119895120579119866(11989611198962)

(2)

where 1198961

= minus1198721 119872

1 1198962

= minus1198722 119872

2 1198821198731

=

119890minus119895(2120587119873

1) 1198821198732

= 119890minus119895(2120587119873

2) and the operator sum

11989911198992

denotessum1198721

1198991=minus1198721

sum1198722

1198992=minus1198722

119860119865(11989611198962)and 119860

119866(11989611198962)are amplitude com-

ponents and 119890119895120579119865(1198961 1198962) and 119890

119895120579119866(1198961 1198962) are phase compo-

nents

01

008

006

004

002

0

600

400

200

0 0200

400

600

800

6

4

2

0

0

400

200200

400

600

8

minus002

Figure 5 POC function for PolyU palmprint database

The cross spectrum 119877(1198961 1198962) between 119865(119896

1 1198962) and

119866(1198961 1198962) is given by

119877 (1198961 1198962) = 119865 (119896

1 1198962) 119866 (119896

1 1198962)

= 119860119865(11989611198962)119860119866(11989611198962)119890119895120579(11989611198962)

(3)

where 119866(1198961 1198962) denotes the complex conjugate of119866(119896

1 1198962)

and 120579(1198961 1198962) = 120579

119865(1198961 1198962) minus 120579119866(1198961 1198962) On the other hand

the cross spectrum (or normalized cross spectrum) (1198961 1198962)

is defined as

(1198961 1198962) =

119865 (1198961 1198962) 119866 (119896

1 1198962)

100381610038161003816100381610038161003816119865 (1198961 1198962) 119866 (119896

1 1198962)100381610038161003816100381610038161003816

= 119890119895120579(11989611198962)

(4)

The POC function 119903119891(1198991 1198992) = is 2D inverse discrete Fourier

transforms (IDFT) of (1198961 1198962) and is given by

119903 (1198991 1198992) =

1

1198731 1198732

(1198961 1198962)119882minus11989611198991

1198731

119882minus11989621198992

1198732

(5)

where sum11989611198962

denotes sum11987211198961=minus1198721

sum1198722

1198962=minus1198722

The most significant property of POC is compared with

the normal correlation and it gives accuracy in patternmatch-ing The distinct sharp peak is obtained when POC functionis the same The peak in the POC drops considerably if thetwo palm print images are not the same Figures 5 6 and 7show the POC function of ROI for PolyU palmprint databaseCOEP palmprint database and IITDelhi palmprint databaseThus the POC function shows much higher perceptionability than the normal correlation function In order tomatch the two palmprint images the height of the peak willbe used as a good resemblance measure

522 Band-Limited Phase-Only Correlation (BLPOC) Inthe POC based image matching method all the frequencycomponents are involved However high frequency com-ponents can be susceptible to noise BLPOC is used toeliminate noisy high frequency components BLPOC bounds

6 Journal of Applied Mathematics

01

008

006

004

002

0

minus002

600

400

200

0 0200

400600

800

0

400

200

0 0200

400600

8

Figure 6 POC function for COEP palmprint database

006

004

002

0

minus002

600

400

200

0 0

200

400

600800

4

2

0

2

0

400

200 400

6008

Figure 7 POC function for IIT Delhi palmprint database

the range of spectrum of the given palmprint image BLPOCfunction between two ROI images can be measured as thePOC function between their low-pass filtered versions basedon the definition of BLPOC Therefore BLPOC functioncan preserve the properties of the POC function BLPOCfunction got a distinct peak if two ROI images are thesame But the locality of peak in two ROI images givesthe translational displacement Experimentations specify thatBLPOC function provides a much higher perception abilitythan the original POC function in palmprint biometricsystem Therefore the efficient size of spectrum is given by1198711= 21198800+ 1 and 119871

2= 21198810+ 1 BLPOC function is given by

11990111988001198810

119892119891(119898 119899) =

1

11987111198712

1198800

sum

119906=minus1198800

1198810

sum

V=minus1198810

119877119866119865(119906V)1198901198952120587(1198981199061198711+119899V1198712) (6)

where 119898 = minus1198800 119880

0 and 119899 = minus119881

0 119881

0 where 119906 =

minus1198800 119880

0 and V = minus119881

0 119881

0 where 0 le 119880

0le 1198720

0 le 1198810le 1198730

Figures 8 9 and 10 are BLPOC function for PolyUpalmprint database COEP palmprint database and IITDelhipalmprint database respectively These examples indicatethat in the instance of genuine matching (a matching per-formed between a pair of palmprint images from the samehand) the BLPOC will show a much sharper peak than POC

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 8 BLPOC function for PolyU palmprint database

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 9 BLPOC function for COEP palmprint database

015

01

005

0

minus005

300

200

100

0 0100

200

300400200

100 200

300

Figure 10 BLPOC function for IIT Delhi palmprint database

however for an imposter matching (a matching performedbetween a couple of palmprint images from different hands)neither BLPOC nor POC will display a different severe peakBLPOC function can be adopted to align the translationbetween two palmprint ROI images and then to measure theresemblance between Fourier transforms of the aligned ROIs

Journal of Applied Mathematics 7

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 11 ROC curve using local features of proposed DOST forPolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 12 ROC curve using global features of proposed DOST forPolyU palmprint database

6 Local-Global Information Fusion (LGIF) forPalmprint Recognition

LGIF based palmprint recognition algorithm is described inthis section The palmprint image acquisition device and theROI extraction systemcandecrease the symmetrical transfor-mations between intraclass ROIs But it is quiet unavoidablebecause there is some translation between intraclass ROIsin the palmprint images These will deteriorate the genuinematching scores This problem is solved by transformingthe given information in horizontal and vertical directionsnumerous times and the smallest of the subsequent similardistances is measured as the final matching distance Thisproblem can be solved in a dissimilar way by calculating thedisplacement parameters between two ROIs using BLPOC

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 13 ROC curve using local features of proposed DOST forCOEP palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 14 ROC curve using global features of proposed DOST forCOEP palmprint database

function Then the mutual regions can be cropped based onwhich the pattern matching is completed The displacementparameters between two palmprint ROI images can be calcu-lated from the peak locality of global BLPOC function Thentwo palmprint ROI images can be aligned based on the dis-placement parameters and extract the mutual regions Ratiobetween the mutual region area and the area of the originalROI can be checked in the palmprint system After alignmentand common area cropping two palmprint ROI images areconstructedThenmatching distance is obtained bymatchingtwo palmprint ROI images The angular distance based onthe normalized Hamming distance is used to match the twopalmprint images for local features in palmprint biometricsystems The two palmprint images 119862

119891and 119862

119892are created

after alignment and common area cropping The matching

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

2 Journal of Applied Mathematics

Image acquisition Preprocessing Local and global feature extraction

Enrollverify

identify

Decision Matching

Database

Figure 1 Block diagram of the proposed palmprint based recognition system

regions The global features can be extracted more consis-tently from the poor quality of the partial palmprints It istoo space and time consuming to directly store and comparethe features pixel by pixel Global feature based matchingoverlaps two given templates with different transformationparameters and estimates the similarity score between thecorresponding cells Compared to local features the globalfeatures have less distinctness and hence been often exploitedtogether with other features or in the preprocessing stageof palmprint matching Both the local and global featuresare fairly independent and capture current information andhence it is realistic to improve the discriminating ability ofmatching by fusing features The local and global features arecombined in the matching stage with accessible feature levelfusing strategies The combined local and global features canimprove the performance of the palmprint systems on largescale databases How to select features is pivotal for the effectof feature combinationThe combination of irrelative featureswill improve the accuracy or efficiency of the palmprintbiometric systems The suitable hierarchical approach canbe employed to reduce the additional time or memory costrequired while fusing the local and global features

In this work a local-global feature fusion technique forpalmprint biometric system is proposed The instantaneousphase information is extracted as the local feature by DOSTThe global feature can be obtained by aggregating the scaleof the DOST The DOST can be reduced to the Fouriertransform of the whole image if the scale of the DOST isincreased to infinity Hence the local features cannot beobtained but the finest resolution for the global frequencyanalysis of the image can be obtained Therefore Fouriertransform coefficients are obviously taken as the globalfeaturesThe arrangement between intraclass palmprint ROIscan also be refined by the global Fourier features The twomatching distances are calculated by matching the local andthe global information separately Then the two matchingdistances are merged by fusion rule to get the final matching

distanceThe block diagram of the proposed system is shownin Figure 1

The rest of this paper is organized as follows Section 2describes the review on palmprint authentication Section 3describes the overview of discrete orthonormal Stockwelltransform (DOST) Preprocessing and ROI extraction isdiscussed in Section 4 Section 5 illustrates the local andglobal feature extraction and matching Section 6 presentslocal-global information fusion for palmprint recognitionExperimental results and discussion are demonstrated inSection 7 Section 8 reports the conclusion

2 Literature Review

In [4] a new BioHashing technique is proposed to generatea cancellable template in the context of unordered set ofnoisyminutiae features and shows that the work is efficient inrealistic context In [17] an intersession variability modelling(ISV) and joint factor analysis (JFA) for face authenticationare proposed to achieve a least significant error rate reductionfor measuring accuracy The line features in palmprint areextracted by using sobel and morphological methods [10]In [18] a new chaff point generation technique for the fuzzyvault in biocryptosystems is proposed and shows that theproposed technique achieves more chaff points compared toKhalil-Hanirsquos algorithm In [19] the principle lines alongwithisolated points are used as features The ridges containingcreases of the palmprint are eliminated and it has beenproposed in [20] The features like end points of principlelines have been used in [9] The end points in the palmprintare found to be directional invariant The crease points thatare related to diseases of a person have been suggested inthe palmprint biometrics system The directional line energyfeatures are also used to identify the palmprint of a personFurther the palmprint biometric systems that are basedon structural features are not invariant to occlusion of thepalmprint image

Journal of Applied Mathematics 3

The feature extraction and patternmatching of palmprintimages are focused mainly The features in the palmprintimages can be classified as local and global features basedon the range of pixels involved in the feature extraction Thelocal feature is a quantity calculatedwithin the local patch andscrambling the detailed behaviours within this specific areaGlobal information is obtained from all or most of the pixelsin the palmprint image and the universal characteristics of theobserved image can be examined Hence palmprint biomet-rics systems can be classified into local based approaches andglobal based approaches according to these definitions Thelocal-global combination for palmprint recognition has beendiscussed only in few papers The local and global featuresare critical for the image observation and identification ofpersons as per the theories in the field of psychophysicsand neurophysiology [22] A global feature reproduces theuniversal features of the image and is appropriate for roughrepresentation A local feature scrambles the more completeinformation within the specific region Therefore local fea-ture is suitable for better representation Hence the improvedrecognition accuracy is obtained if the local and global featurecan be appropriately combined It is already used in biomet-rics systems like iris face and fingerprint The conventionalDaugman-like classifier uses global features enclosed by zero-crossing boundaries in the palmprint biometrics systemsIn two-level palmprint matching scheme [23] the globalfeatures for coarse level filtering can be extracted by Houghtransform and for fine-level matching the local featuresextracted from the localities and directions of the principallines in the palmprint

The local and global features are obtained from thenonnegative matrix factorization technique and principlecomponent analysis in [24] can be combined for palmprintbiometric system In face recognition [6] the global andlocal information are extracted by principle componentanalysis (PCA) and Haar wavelets Global features in the faceimages are extracted by keeping the low-frequency Fouriercoefficients and the local information is extracted usingGabor filters The local and global information combinationwas also utilized in the fingerprint biometrics system [25]

3 Overview of Discrete OrthonormalStockwell Transform

The S-transform is more powerful than other multiresolutiontechniques like short time Fourier transform (STFT) andwavelet transform (WT) The phase of the S-transform refer-enced to the time origin provides useful and supplementaryinformation about spectra that is not available from locallyreferenced phase information in the continuous wavelettransform (CWT)

The S-transform is advantageous for the analysis ofpalmprint images as it preserves the phase information usinglinear frequency scaling

(1) However the major limitation of S-transform is itshigh time and space complexity due to its redundantnature which makes it impractical in many cases

(2) The 2D-ST of an array of size119873 times 119873 has a computa-tional complexity of O (119873

4+ 1198734 log119873) and storage

requirements of O (1198734)(3) To eliminate this problem of 2D-ST we use DOST

which is also a multiresolution technique for extrac-tion of features from the palmprint images and isbased on the S-transform

(4) DOST have less computational and storage complex-ity in comparison to the S-transform because it usesan orthonormal set of basis functions while retainingall the advantageous properties of S-transform

(5) 2D-DOSTprovides a spatial frequency representationof an image with computational and storage com-plexity asO (1198732+1198732 log119873) andO (1198732) respectively

With the dyadic sampling scheme in order 0 1 2 log2119873minus

1 DOST of an119873 times119873 palmprint image 119891(119909 119910) is performedby the following steps

(1) Two-dimensional Fourier transform (FT) is appliedto the image119891(119909 119910) to obtain Fourier samples119865(119906 V)

(2) Partition the Fourier sample 119865(119906 V) and multiply itby the square root of the number of points in thepartition and perform an inverse FT Then the voiceimage is calculated as

119878 (1199091015840 1199101015840 V119909 V119910) =

1

radic2119875119909119875119910minus2

2119875

119909minus2minus1

sum

119906=minus2119901119909minus2

2119875

119910minus2

minus1

sum

V=minus2119875119910minus2119865 (119906 + V

119909 V + V

119910)

here V119909= 2119901119909minus1+2119875119909minus2

V119910= 2119901119910minus1+2119875119910minus2

1198902120587119894

(1199061199091015840

2119875119909minus1+

V1199101015840

2119875119910minus1)

(1)

where V119909and V119910are the horizontal and vertical voice

frequencies where 1199091015840 corresponds to 1199091015840-coordinate

in space where 1199101015840 corresponds to 1199101015840-coordinate in

space(3) Thus the features of the palmprint images are

obtained after the transformation

4 Preprocessing and ROI Extraction

In this section the acquired hand image is preprocessed andthe palmprint region is extracted The square area insidethe palm region of the hand image is considered as thepalmprint or region of interest (ROI) Normally the extractedpalm region has the rotation and translation error In orderto minimize the rotation and translation error the ROIalgorithm identifies gaps between fingers The gaps betweenthe fingers are used as the reference points to align theimage The images of the extracted palm region are alignedrotationally by levelling the gap between the index finger

4 Journal of Applied Mathematics

(a) (b)

Figure 2 (a) Sample input image (b) ROI image for PolyU palmprint database

(a) (b)

Figure 3 (a) Sample input image (b) ROI image for COEP palmprint database

(a) (b)

Figure 4 (a) Sample input image (b) ROI image for IIT Delhi palmprint database

(IF) and the middle finger (MF) and that between the MFand the ring finger (RF) Among the three gaps the gapbetween MF and RF is the toughest point In order toeliminate the translation error the gap between MF and RFis used as the reference point The maximum square regionextracted from the palm area is selected as the ROI Thesquare region is horizontally centered on the axis runningthrough the gap between MF and RF Due to the regular andcontrolled uniform illumination conditions during imagecapture the acquired hand image and its background contrastin colour The sample input image is shown in Figures 2(a)

3(a) and 4(a) for PolyU palmprint database COEP palm-print database and IIT Delhi palmprint database Globalthresholding has been applied to extract the hand from thebackgroundThe image is processedwith opening and closingmorphological operations to remove any isolated small blobsor holes Contour-tracing algorithm is applied to extract thecontour of the hand image from the palmprint The ROIextraction technique of IIT Delhi palmprint database PolyUpalmprint database and COEP palmprint database are sameas in [9 26] The corresponding Figures 2(b) 3(b) and 4(b)are shown in this paper which gives the details of ROI image

Journal of Applied Mathematics 5

5 Local-Global FeatureExtraction and Matching

51 Local Feature The feature extraction technique in anypalmprint biometric system is used to obtain good interclassseparation in minimum time The features of palmprintare extracted by using 2D-DOST after the completion ofpreprocessing and ROI segmentation from palmprint Thelocal variation of instantaneous phase is used to extractfeatures from palmprint Instantaneous phase obtained usingDOST is the resolution of phase with respective to time Ithas the merits of great accurateness sturdiness to brightnessvariation and fast matching When matching two palmprintimages the angular distance based on the normalized Ham-ming distance is used

52 Global Feature DOST is used to extract the instanta-neous phase information and also it can be considered as awindowed Fourier transform The DOST can be consideredas a windowed Fourier transformThe phase-only correlation(POC) and band-limited phase-only correlation (BLPOC)are utilized to match the global information of the twopalmprint images

521 Phase-Only Correlation (POC) The Fourier transformcoefficients are considered as the global information Thetwo given Fourier transforms are matched using POC POCbased techniques have been widely used in image registrationtasks Recently a POC technique considered as a resemblancemeasure in certain biometrics systems [7 22] BLPOC ismore effective when compared to the conventional POCBLPOC is utilized to estimate the translation parametersbetween palmprint ROIs and also measure the resemblanceof the Fourier transforms of the aligned ROIs POC is anoperational method to estimate the displacement parametersbetween two palmprint images in the Fourier transformdomain Its primary principle is the translation property ofthe Fourier transforms

Consider two 1198731times 1198732images 119891(119899

1 1198992) and 119892(119899

1 1198992)

where the assumption is that the index ranges are 1198991

=

minus1198721 119872

1(1198721gt 0) and 119899

2= minus119872

2 119872

2(1198722gt 0)

For mathematical simplicity 1198731

= 21198721+ 1 and 119873

2=

21198722+ 1 Let 119865(119896

1 1198962) and 119866(119896

1 1198962) denote the 2D discrete

Fourier transforms (2D DFTs) of the two images 119865(1198961 1198962)

and 119866(1198961 1198962) are given by

119865 (1198961 1198962) = sum

11989911198992

119891 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119865(11989611198962)119890119895120579119865(11989611198962)

119866 (1198961 1198962) = sum

11989911198992

119892 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119866(11989611198962)119890119895120579119866(11989611198962)

(2)

where 1198961

= minus1198721 119872

1 1198962

= minus1198722 119872

2 1198821198731

=

119890minus119895(2120587119873

1) 1198821198732

= 119890minus119895(2120587119873

2) and the operator sum

11989911198992

denotessum1198721

1198991=minus1198721

sum1198722

1198992=minus1198722

119860119865(11989611198962)and 119860

119866(11989611198962)are amplitude com-

ponents and 119890119895120579119865(1198961 1198962) and 119890

119895120579119866(1198961 1198962) are phase compo-

nents

01

008

006

004

002

0

600

400

200

0 0200

400

600

800

6

4

2

0

0

400

200200

400

600

8

minus002

Figure 5 POC function for PolyU palmprint database

The cross spectrum 119877(1198961 1198962) between 119865(119896

1 1198962) and

119866(1198961 1198962) is given by

119877 (1198961 1198962) = 119865 (119896

1 1198962) 119866 (119896

1 1198962)

= 119860119865(11989611198962)119860119866(11989611198962)119890119895120579(11989611198962)

(3)

where 119866(1198961 1198962) denotes the complex conjugate of119866(119896

1 1198962)

and 120579(1198961 1198962) = 120579

119865(1198961 1198962) minus 120579119866(1198961 1198962) On the other hand

the cross spectrum (or normalized cross spectrum) (1198961 1198962)

is defined as

(1198961 1198962) =

119865 (1198961 1198962) 119866 (119896

1 1198962)

100381610038161003816100381610038161003816119865 (1198961 1198962) 119866 (119896

1 1198962)100381610038161003816100381610038161003816

= 119890119895120579(11989611198962)

(4)

The POC function 119903119891(1198991 1198992) = is 2D inverse discrete Fourier

transforms (IDFT) of (1198961 1198962) and is given by

119903 (1198991 1198992) =

1

1198731 1198732

(1198961 1198962)119882minus11989611198991

1198731

119882minus11989621198992

1198732

(5)

where sum11989611198962

denotes sum11987211198961=minus1198721

sum1198722

1198962=minus1198722

The most significant property of POC is compared with

the normal correlation and it gives accuracy in patternmatch-ing The distinct sharp peak is obtained when POC functionis the same The peak in the POC drops considerably if thetwo palm print images are not the same Figures 5 6 and 7show the POC function of ROI for PolyU palmprint databaseCOEP palmprint database and IITDelhi palmprint databaseThus the POC function shows much higher perceptionability than the normal correlation function In order tomatch the two palmprint images the height of the peak willbe used as a good resemblance measure

522 Band-Limited Phase-Only Correlation (BLPOC) Inthe POC based image matching method all the frequencycomponents are involved However high frequency com-ponents can be susceptible to noise BLPOC is used toeliminate noisy high frequency components BLPOC bounds

6 Journal of Applied Mathematics

01

008

006

004

002

0

minus002

600

400

200

0 0200

400600

800

0

400

200

0 0200

400600

8

Figure 6 POC function for COEP palmprint database

006

004

002

0

minus002

600

400

200

0 0

200

400

600800

4

2

0

2

0

400

200 400

6008

Figure 7 POC function for IIT Delhi palmprint database

the range of spectrum of the given palmprint image BLPOCfunction between two ROI images can be measured as thePOC function between their low-pass filtered versions basedon the definition of BLPOC Therefore BLPOC functioncan preserve the properties of the POC function BLPOCfunction got a distinct peak if two ROI images are thesame But the locality of peak in two ROI images givesthe translational displacement Experimentations specify thatBLPOC function provides a much higher perception abilitythan the original POC function in palmprint biometricsystem Therefore the efficient size of spectrum is given by1198711= 21198800+ 1 and 119871

2= 21198810+ 1 BLPOC function is given by

11990111988001198810

119892119891(119898 119899) =

1

11987111198712

1198800

sum

119906=minus1198800

1198810

sum

V=minus1198810

119877119866119865(119906V)1198901198952120587(1198981199061198711+119899V1198712) (6)

where 119898 = minus1198800 119880

0 and 119899 = minus119881

0 119881

0 where 119906 =

minus1198800 119880

0 and V = minus119881

0 119881

0 where 0 le 119880

0le 1198720

0 le 1198810le 1198730

Figures 8 9 and 10 are BLPOC function for PolyUpalmprint database COEP palmprint database and IITDelhipalmprint database respectively These examples indicatethat in the instance of genuine matching (a matching per-formed between a pair of palmprint images from the samehand) the BLPOC will show a much sharper peak than POC

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 8 BLPOC function for PolyU palmprint database

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 9 BLPOC function for COEP palmprint database

015

01

005

0

minus005

300

200

100

0 0100

200

300400200

100 200

300

Figure 10 BLPOC function for IIT Delhi palmprint database

however for an imposter matching (a matching performedbetween a couple of palmprint images from different hands)neither BLPOC nor POC will display a different severe peakBLPOC function can be adopted to align the translationbetween two palmprint ROI images and then to measure theresemblance between Fourier transforms of the aligned ROIs

Journal of Applied Mathematics 7

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 11 ROC curve using local features of proposed DOST forPolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 12 ROC curve using global features of proposed DOST forPolyU palmprint database

6 Local-Global Information Fusion (LGIF) forPalmprint Recognition

LGIF based palmprint recognition algorithm is described inthis section The palmprint image acquisition device and theROI extraction systemcandecrease the symmetrical transfor-mations between intraclass ROIs But it is quiet unavoidablebecause there is some translation between intraclass ROIsin the palmprint images These will deteriorate the genuinematching scores This problem is solved by transformingthe given information in horizontal and vertical directionsnumerous times and the smallest of the subsequent similardistances is measured as the final matching distance Thisproblem can be solved in a dissimilar way by calculating thedisplacement parameters between two ROIs using BLPOC

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 13 ROC curve using local features of proposed DOST forCOEP palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 14 ROC curve using global features of proposed DOST forCOEP palmprint database

function Then the mutual regions can be cropped based onwhich the pattern matching is completed The displacementparameters between two palmprint ROI images can be calcu-lated from the peak locality of global BLPOC function Thentwo palmprint ROI images can be aligned based on the dis-placement parameters and extract the mutual regions Ratiobetween the mutual region area and the area of the originalROI can be checked in the palmprint system After alignmentand common area cropping two palmprint ROI images areconstructedThenmatching distance is obtained bymatchingtwo palmprint ROI images The angular distance based onthe normalized Hamming distance is used to match the twopalmprint images for local features in palmprint biometricsystems The two palmprint images 119862

119891and 119862

119892are created

after alignment and common area cropping The matching

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Journal of Applied Mathematics 3

The feature extraction and patternmatching of palmprintimages are focused mainly The features in the palmprintimages can be classified as local and global features basedon the range of pixels involved in the feature extraction Thelocal feature is a quantity calculatedwithin the local patch andscrambling the detailed behaviours within this specific areaGlobal information is obtained from all or most of the pixelsin the palmprint image and the universal characteristics of theobserved image can be examined Hence palmprint biomet-rics systems can be classified into local based approaches andglobal based approaches according to these definitions Thelocal-global combination for palmprint recognition has beendiscussed only in few papers The local and global featuresare critical for the image observation and identification ofpersons as per the theories in the field of psychophysicsand neurophysiology [22] A global feature reproduces theuniversal features of the image and is appropriate for roughrepresentation A local feature scrambles the more completeinformation within the specific region Therefore local fea-ture is suitable for better representation Hence the improvedrecognition accuracy is obtained if the local and global featurecan be appropriately combined It is already used in biomet-rics systems like iris face and fingerprint The conventionalDaugman-like classifier uses global features enclosed by zero-crossing boundaries in the palmprint biometrics systemsIn two-level palmprint matching scheme [23] the globalfeatures for coarse level filtering can be extracted by Houghtransform and for fine-level matching the local featuresextracted from the localities and directions of the principallines in the palmprint

The local and global features are obtained from thenonnegative matrix factorization technique and principlecomponent analysis in [24] can be combined for palmprintbiometric system In face recognition [6] the global andlocal information are extracted by principle componentanalysis (PCA) and Haar wavelets Global features in the faceimages are extracted by keeping the low-frequency Fouriercoefficients and the local information is extracted usingGabor filters The local and global information combinationwas also utilized in the fingerprint biometrics system [25]

3 Overview of Discrete OrthonormalStockwell Transform

The S-transform is more powerful than other multiresolutiontechniques like short time Fourier transform (STFT) andwavelet transform (WT) The phase of the S-transform refer-enced to the time origin provides useful and supplementaryinformation about spectra that is not available from locallyreferenced phase information in the continuous wavelettransform (CWT)

The S-transform is advantageous for the analysis ofpalmprint images as it preserves the phase information usinglinear frequency scaling

(1) However the major limitation of S-transform is itshigh time and space complexity due to its redundantnature which makes it impractical in many cases

(2) The 2D-ST of an array of size119873 times 119873 has a computa-tional complexity of O (119873

4+ 1198734 log119873) and storage

requirements of O (1198734)(3) To eliminate this problem of 2D-ST we use DOST

which is also a multiresolution technique for extrac-tion of features from the palmprint images and isbased on the S-transform

(4) DOST have less computational and storage complex-ity in comparison to the S-transform because it usesan orthonormal set of basis functions while retainingall the advantageous properties of S-transform

(5) 2D-DOSTprovides a spatial frequency representationof an image with computational and storage com-plexity asO (1198732+1198732 log119873) andO (1198732) respectively

With the dyadic sampling scheme in order 0 1 2 log2119873minus

1 DOST of an119873 times119873 palmprint image 119891(119909 119910) is performedby the following steps

(1) Two-dimensional Fourier transform (FT) is appliedto the image119891(119909 119910) to obtain Fourier samples119865(119906 V)

(2) Partition the Fourier sample 119865(119906 V) and multiply itby the square root of the number of points in thepartition and perform an inverse FT Then the voiceimage is calculated as

119878 (1199091015840 1199101015840 V119909 V119910) =

1

radic2119875119909119875119910minus2

2119875

119909minus2minus1

sum

119906=minus2119901119909minus2

2119875

119910minus2

minus1

sum

V=minus2119875119910minus2119865 (119906 + V

119909 V + V

119910)

here V119909= 2119901119909minus1+2119875119909minus2

V119910= 2119901119910minus1+2119875119910minus2

1198902120587119894

(1199061199091015840

2119875119909minus1+

V1199101015840

2119875119910minus1)

(1)

where V119909and V119910are the horizontal and vertical voice

frequencies where 1199091015840 corresponds to 1199091015840-coordinate

in space where 1199101015840 corresponds to 1199101015840-coordinate in

space(3) Thus the features of the palmprint images are

obtained after the transformation

4 Preprocessing and ROI Extraction

In this section the acquired hand image is preprocessed andthe palmprint region is extracted The square area insidethe palm region of the hand image is considered as thepalmprint or region of interest (ROI) Normally the extractedpalm region has the rotation and translation error In orderto minimize the rotation and translation error the ROIalgorithm identifies gaps between fingers The gaps betweenthe fingers are used as the reference points to align theimage The images of the extracted palm region are alignedrotationally by levelling the gap between the index finger

4 Journal of Applied Mathematics

(a) (b)

Figure 2 (a) Sample input image (b) ROI image for PolyU palmprint database

(a) (b)

Figure 3 (a) Sample input image (b) ROI image for COEP palmprint database

(a) (b)

Figure 4 (a) Sample input image (b) ROI image for IIT Delhi palmprint database

(IF) and the middle finger (MF) and that between the MFand the ring finger (RF) Among the three gaps the gapbetween MF and RF is the toughest point In order toeliminate the translation error the gap between MF and RFis used as the reference point The maximum square regionextracted from the palm area is selected as the ROI Thesquare region is horizontally centered on the axis runningthrough the gap between MF and RF Due to the regular andcontrolled uniform illumination conditions during imagecapture the acquired hand image and its background contrastin colour The sample input image is shown in Figures 2(a)

3(a) and 4(a) for PolyU palmprint database COEP palm-print database and IIT Delhi palmprint database Globalthresholding has been applied to extract the hand from thebackgroundThe image is processedwith opening and closingmorphological operations to remove any isolated small blobsor holes Contour-tracing algorithm is applied to extract thecontour of the hand image from the palmprint The ROIextraction technique of IIT Delhi palmprint database PolyUpalmprint database and COEP palmprint database are sameas in [9 26] The corresponding Figures 2(b) 3(b) and 4(b)are shown in this paper which gives the details of ROI image

Journal of Applied Mathematics 5

5 Local-Global FeatureExtraction and Matching

51 Local Feature The feature extraction technique in anypalmprint biometric system is used to obtain good interclassseparation in minimum time The features of palmprintare extracted by using 2D-DOST after the completion ofpreprocessing and ROI segmentation from palmprint Thelocal variation of instantaneous phase is used to extractfeatures from palmprint Instantaneous phase obtained usingDOST is the resolution of phase with respective to time Ithas the merits of great accurateness sturdiness to brightnessvariation and fast matching When matching two palmprintimages the angular distance based on the normalized Ham-ming distance is used

52 Global Feature DOST is used to extract the instanta-neous phase information and also it can be considered as awindowed Fourier transform The DOST can be consideredas a windowed Fourier transformThe phase-only correlation(POC) and band-limited phase-only correlation (BLPOC)are utilized to match the global information of the twopalmprint images

521 Phase-Only Correlation (POC) The Fourier transformcoefficients are considered as the global information Thetwo given Fourier transforms are matched using POC POCbased techniques have been widely used in image registrationtasks Recently a POC technique considered as a resemblancemeasure in certain biometrics systems [7 22] BLPOC ismore effective when compared to the conventional POCBLPOC is utilized to estimate the translation parametersbetween palmprint ROIs and also measure the resemblanceof the Fourier transforms of the aligned ROIs POC is anoperational method to estimate the displacement parametersbetween two palmprint images in the Fourier transformdomain Its primary principle is the translation property ofthe Fourier transforms

Consider two 1198731times 1198732images 119891(119899

1 1198992) and 119892(119899

1 1198992)

where the assumption is that the index ranges are 1198991

=

minus1198721 119872

1(1198721gt 0) and 119899

2= minus119872

2 119872

2(1198722gt 0)

For mathematical simplicity 1198731

= 21198721+ 1 and 119873

2=

21198722+ 1 Let 119865(119896

1 1198962) and 119866(119896

1 1198962) denote the 2D discrete

Fourier transforms (2D DFTs) of the two images 119865(1198961 1198962)

and 119866(1198961 1198962) are given by

119865 (1198961 1198962) = sum

11989911198992

119891 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119865(11989611198962)119890119895120579119865(11989611198962)

119866 (1198961 1198962) = sum

11989911198992

119892 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119866(11989611198962)119890119895120579119866(11989611198962)

(2)

where 1198961

= minus1198721 119872

1 1198962

= minus1198722 119872

2 1198821198731

=

119890minus119895(2120587119873

1) 1198821198732

= 119890minus119895(2120587119873

2) and the operator sum

11989911198992

denotessum1198721

1198991=minus1198721

sum1198722

1198992=minus1198722

119860119865(11989611198962)and 119860

119866(11989611198962)are amplitude com-

ponents and 119890119895120579119865(1198961 1198962) and 119890

119895120579119866(1198961 1198962) are phase compo-

nents

01

008

006

004

002

0

600

400

200

0 0200

400

600

800

6

4

2

0

0

400

200200

400

600

8

minus002

Figure 5 POC function for PolyU palmprint database

The cross spectrum 119877(1198961 1198962) between 119865(119896

1 1198962) and

119866(1198961 1198962) is given by

119877 (1198961 1198962) = 119865 (119896

1 1198962) 119866 (119896

1 1198962)

= 119860119865(11989611198962)119860119866(11989611198962)119890119895120579(11989611198962)

(3)

where 119866(1198961 1198962) denotes the complex conjugate of119866(119896

1 1198962)

and 120579(1198961 1198962) = 120579

119865(1198961 1198962) minus 120579119866(1198961 1198962) On the other hand

the cross spectrum (or normalized cross spectrum) (1198961 1198962)

is defined as

(1198961 1198962) =

119865 (1198961 1198962) 119866 (119896

1 1198962)

100381610038161003816100381610038161003816119865 (1198961 1198962) 119866 (119896

1 1198962)100381610038161003816100381610038161003816

= 119890119895120579(11989611198962)

(4)

The POC function 119903119891(1198991 1198992) = is 2D inverse discrete Fourier

transforms (IDFT) of (1198961 1198962) and is given by

119903 (1198991 1198992) =

1

1198731 1198732

(1198961 1198962)119882minus11989611198991

1198731

119882minus11989621198992

1198732

(5)

where sum11989611198962

denotes sum11987211198961=minus1198721

sum1198722

1198962=minus1198722

The most significant property of POC is compared with

the normal correlation and it gives accuracy in patternmatch-ing The distinct sharp peak is obtained when POC functionis the same The peak in the POC drops considerably if thetwo palm print images are not the same Figures 5 6 and 7show the POC function of ROI for PolyU palmprint databaseCOEP palmprint database and IITDelhi palmprint databaseThus the POC function shows much higher perceptionability than the normal correlation function In order tomatch the two palmprint images the height of the peak willbe used as a good resemblance measure

522 Band-Limited Phase-Only Correlation (BLPOC) Inthe POC based image matching method all the frequencycomponents are involved However high frequency com-ponents can be susceptible to noise BLPOC is used toeliminate noisy high frequency components BLPOC bounds

6 Journal of Applied Mathematics

01

008

006

004

002

0

minus002

600

400

200

0 0200

400600

800

0

400

200

0 0200

400600

8

Figure 6 POC function for COEP palmprint database

006

004

002

0

minus002

600

400

200

0 0

200

400

600800

4

2

0

2

0

400

200 400

6008

Figure 7 POC function for IIT Delhi palmprint database

the range of spectrum of the given palmprint image BLPOCfunction between two ROI images can be measured as thePOC function between their low-pass filtered versions basedon the definition of BLPOC Therefore BLPOC functioncan preserve the properties of the POC function BLPOCfunction got a distinct peak if two ROI images are thesame But the locality of peak in two ROI images givesthe translational displacement Experimentations specify thatBLPOC function provides a much higher perception abilitythan the original POC function in palmprint biometricsystem Therefore the efficient size of spectrum is given by1198711= 21198800+ 1 and 119871

2= 21198810+ 1 BLPOC function is given by

11990111988001198810

119892119891(119898 119899) =

1

11987111198712

1198800

sum

119906=minus1198800

1198810

sum

V=minus1198810

119877119866119865(119906V)1198901198952120587(1198981199061198711+119899V1198712) (6)

where 119898 = minus1198800 119880

0 and 119899 = minus119881

0 119881

0 where 119906 =

minus1198800 119880

0 and V = minus119881

0 119881

0 where 0 le 119880

0le 1198720

0 le 1198810le 1198730

Figures 8 9 and 10 are BLPOC function for PolyUpalmprint database COEP palmprint database and IITDelhipalmprint database respectively These examples indicatethat in the instance of genuine matching (a matching per-formed between a pair of palmprint images from the samehand) the BLPOC will show a much sharper peak than POC

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 8 BLPOC function for PolyU palmprint database

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 9 BLPOC function for COEP palmprint database

015

01

005

0

minus005

300

200

100

0 0100

200

300400200

100 200

300

Figure 10 BLPOC function for IIT Delhi palmprint database

however for an imposter matching (a matching performedbetween a couple of palmprint images from different hands)neither BLPOC nor POC will display a different severe peakBLPOC function can be adopted to align the translationbetween two palmprint ROI images and then to measure theresemblance between Fourier transforms of the aligned ROIs

Journal of Applied Mathematics 7

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 11 ROC curve using local features of proposed DOST forPolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 12 ROC curve using global features of proposed DOST forPolyU palmprint database

6 Local-Global Information Fusion (LGIF) forPalmprint Recognition

LGIF based palmprint recognition algorithm is described inthis section The palmprint image acquisition device and theROI extraction systemcandecrease the symmetrical transfor-mations between intraclass ROIs But it is quiet unavoidablebecause there is some translation between intraclass ROIsin the palmprint images These will deteriorate the genuinematching scores This problem is solved by transformingthe given information in horizontal and vertical directionsnumerous times and the smallest of the subsequent similardistances is measured as the final matching distance Thisproblem can be solved in a dissimilar way by calculating thedisplacement parameters between two ROIs using BLPOC

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 13 ROC curve using local features of proposed DOST forCOEP palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 14 ROC curve using global features of proposed DOST forCOEP palmprint database

function Then the mutual regions can be cropped based onwhich the pattern matching is completed The displacementparameters between two palmprint ROI images can be calcu-lated from the peak locality of global BLPOC function Thentwo palmprint ROI images can be aligned based on the dis-placement parameters and extract the mutual regions Ratiobetween the mutual region area and the area of the originalROI can be checked in the palmprint system After alignmentand common area cropping two palmprint ROI images areconstructedThenmatching distance is obtained bymatchingtwo palmprint ROI images The angular distance based onthe normalized Hamming distance is used to match the twopalmprint images for local features in palmprint biometricsystems The two palmprint images 119862

119891and 119862

119892are created

after alignment and common area cropping The matching

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

4 Journal of Applied Mathematics

(a) (b)

Figure 2 (a) Sample input image (b) ROI image for PolyU palmprint database

(a) (b)

Figure 3 (a) Sample input image (b) ROI image for COEP palmprint database

(a) (b)

Figure 4 (a) Sample input image (b) ROI image for IIT Delhi palmprint database

(IF) and the middle finger (MF) and that between the MFand the ring finger (RF) Among the three gaps the gapbetween MF and RF is the toughest point In order toeliminate the translation error the gap between MF and RFis used as the reference point The maximum square regionextracted from the palm area is selected as the ROI Thesquare region is horizontally centered on the axis runningthrough the gap between MF and RF Due to the regular andcontrolled uniform illumination conditions during imagecapture the acquired hand image and its background contrastin colour The sample input image is shown in Figures 2(a)

3(a) and 4(a) for PolyU palmprint database COEP palm-print database and IIT Delhi palmprint database Globalthresholding has been applied to extract the hand from thebackgroundThe image is processedwith opening and closingmorphological operations to remove any isolated small blobsor holes Contour-tracing algorithm is applied to extract thecontour of the hand image from the palmprint The ROIextraction technique of IIT Delhi palmprint database PolyUpalmprint database and COEP palmprint database are sameas in [9 26] The corresponding Figures 2(b) 3(b) and 4(b)are shown in this paper which gives the details of ROI image

Journal of Applied Mathematics 5

5 Local-Global FeatureExtraction and Matching

51 Local Feature The feature extraction technique in anypalmprint biometric system is used to obtain good interclassseparation in minimum time The features of palmprintare extracted by using 2D-DOST after the completion ofpreprocessing and ROI segmentation from palmprint Thelocal variation of instantaneous phase is used to extractfeatures from palmprint Instantaneous phase obtained usingDOST is the resolution of phase with respective to time Ithas the merits of great accurateness sturdiness to brightnessvariation and fast matching When matching two palmprintimages the angular distance based on the normalized Ham-ming distance is used

52 Global Feature DOST is used to extract the instanta-neous phase information and also it can be considered as awindowed Fourier transform The DOST can be consideredas a windowed Fourier transformThe phase-only correlation(POC) and band-limited phase-only correlation (BLPOC)are utilized to match the global information of the twopalmprint images

521 Phase-Only Correlation (POC) The Fourier transformcoefficients are considered as the global information Thetwo given Fourier transforms are matched using POC POCbased techniques have been widely used in image registrationtasks Recently a POC technique considered as a resemblancemeasure in certain biometrics systems [7 22] BLPOC ismore effective when compared to the conventional POCBLPOC is utilized to estimate the translation parametersbetween palmprint ROIs and also measure the resemblanceof the Fourier transforms of the aligned ROIs POC is anoperational method to estimate the displacement parametersbetween two palmprint images in the Fourier transformdomain Its primary principle is the translation property ofthe Fourier transforms

Consider two 1198731times 1198732images 119891(119899

1 1198992) and 119892(119899

1 1198992)

where the assumption is that the index ranges are 1198991

=

minus1198721 119872

1(1198721gt 0) and 119899

2= minus119872

2 119872

2(1198722gt 0)

For mathematical simplicity 1198731

= 21198721+ 1 and 119873

2=

21198722+ 1 Let 119865(119896

1 1198962) and 119866(119896

1 1198962) denote the 2D discrete

Fourier transforms (2D DFTs) of the two images 119865(1198961 1198962)

and 119866(1198961 1198962) are given by

119865 (1198961 1198962) = sum

11989911198992

119891 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119865(11989611198962)119890119895120579119865(11989611198962)

119866 (1198961 1198962) = sum

11989911198992

119892 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119866(11989611198962)119890119895120579119866(11989611198962)

(2)

where 1198961

= minus1198721 119872

1 1198962

= minus1198722 119872

2 1198821198731

=

119890minus119895(2120587119873

1) 1198821198732

= 119890minus119895(2120587119873

2) and the operator sum

11989911198992

denotessum1198721

1198991=minus1198721

sum1198722

1198992=minus1198722

119860119865(11989611198962)and 119860

119866(11989611198962)are amplitude com-

ponents and 119890119895120579119865(1198961 1198962) and 119890

119895120579119866(1198961 1198962) are phase compo-

nents

01

008

006

004

002

0

600

400

200

0 0200

400

600

800

6

4

2

0

0

400

200200

400

600

8

minus002

Figure 5 POC function for PolyU palmprint database

The cross spectrum 119877(1198961 1198962) between 119865(119896

1 1198962) and

119866(1198961 1198962) is given by

119877 (1198961 1198962) = 119865 (119896

1 1198962) 119866 (119896

1 1198962)

= 119860119865(11989611198962)119860119866(11989611198962)119890119895120579(11989611198962)

(3)

where 119866(1198961 1198962) denotes the complex conjugate of119866(119896

1 1198962)

and 120579(1198961 1198962) = 120579

119865(1198961 1198962) minus 120579119866(1198961 1198962) On the other hand

the cross spectrum (or normalized cross spectrum) (1198961 1198962)

is defined as

(1198961 1198962) =

119865 (1198961 1198962) 119866 (119896

1 1198962)

100381610038161003816100381610038161003816119865 (1198961 1198962) 119866 (119896

1 1198962)100381610038161003816100381610038161003816

= 119890119895120579(11989611198962)

(4)

The POC function 119903119891(1198991 1198992) = is 2D inverse discrete Fourier

transforms (IDFT) of (1198961 1198962) and is given by

119903 (1198991 1198992) =

1

1198731 1198732

(1198961 1198962)119882minus11989611198991

1198731

119882minus11989621198992

1198732

(5)

where sum11989611198962

denotes sum11987211198961=minus1198721

sum1198722

1198962=minus1198722

The most significant property of POC is compared with

the normal correlation and it gives accuracy in patternmatch-ing The distinct sharp peak is obtained when POC functionis the same The peak in the POC drops considerably if thetwo palm print images are not the same Figures 5 6 and 7show the POC function of ROI for PolyU palmprint databaseCOEP palmprint database and IITDelhi palmprint databaseThus the POC function shows much higher perceptionability than the normal correlation function In order tomatch the two palmprint images the height of the peak willbe used as a good resemblance measure

522 Band-Limited Phase-Only Correlation (BLPOC) Inthe POC based image matching method all the frequencycomponents are involved However high frequency com-ponents can be susceptible to noise BLPOC is used toeliminate noisy high frequency components BLPOC bounds

6 Journal of Applied Mathematics

01

008

006

004

002

0

minus002

600

400

200

0 0200

400600

800

0

400

200

0 0200

400600

8

Figure 6 POC function for COEP palmprint database

006

004

002

0

minus002

600

400

200

0 0

200

400

600800

4

2

0

2

0

400

200 400

6008

Figure 7 POC function for IIT Delhi palmprint database

the range of spectrum of the given palmprint image BLPOCfunction between two ROI images can be measured as thePOC function between their low-pass filtered versions basedon the definition of BLPOC Therefore BLPOC functioncan preserve the properties of the POC function BLPOCfunction got a distinct peak if two ROI images are thesame But the locality of peak in two ROI images givesthe translational displacement Experimentations specify thatBLPOC function provides a much higher perception abilitythan the original POC function in palmprint biometricsystem Therefore the efficient size of spectrum is given by1198711= 21198800+ 1 and 119871

2= 21198810+ 1 BLPOC function is given by

11990111988001198810

119892119891(119898 119899) =

1

11987111198712

1198800

sum

119906=minus1198800

1198810

sum

V=minus1198810

119877119866119865(119906V)1198901198952120587(1198981199061198711+119899V1198712) (6)

where 119898 = minus1198800 119880

0 and 119899 = minus119881

0 119881

0 where 119906 =

minus1198800 119880

0 and V = minus119881

0 119881

0 where 0 le 119880

0le 1198720

0 le 1198810le 1198730

Figures 8 9 and 10 are BLPOC function for PolyUpalmprint database COEP palmprint database and IITDelhipalmprint database respectively These examples indicatethat in the instance of genuine matching (a matching per-formed between a pair of palmprint images from the samehand) the BLPOC will show a much sharper peak than POC

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 8 BLPOC function for PolyU palmprint database

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 9 BLPOC function for COEP palmprint database

015

01

005

0

minus005

300

200

100

0 0100

200

300400200

100 200

300

Figure 10 BLPOC function for IIT Delhi palmprint database

however for an imposter matching (a matching performedbetween a couple of palmprint images from different hands)neither BLPOC nor POC will display a different severe peakBLPOC function can be adopted to align the translationbetween two palmprint ROI images and then to measure theresemblance between Fourier transforms of the aligned ROIs

Journal of Applied Mathematics 7

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 11 ROC curve using local features of proposed DOST forPolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 12 ROC curve using global features of proposed DOST forPolyU palmprint database

6 Local-Global Information Fusion (LGIF) forPalmprint Recognition

LGIF based palmprint recognition algorithm is described inthis section The palmprint image acquisition device and theROI extraction systemcandecrease the symmetrical transfor-mations between intraclass ROIs But it is quiet unavoidablebecause there is some translation between intraclass ROIsin the palmprint images These will deteriorate the genuinematching scores This problem is solved by transformingthe given information in horizontal and vertical directionsnumerous times and the smallest of the subsequent similardistances is measured as the final matching distance Thisproblem can be solved in a dissimilar way by calculating thedisplacement parameters between two ROIs using BLPOC

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 13 ROC curve using local features of proposed DOST forCOEP palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 14 ROC curve using global features of proposed DOST forCOEP palmprint database

function Then the mutual regions can be cropped based onwhich the pattern matching is completed The displacementparameters between two palmprint ROI images can be calcu-lated from the peak locality of global BLPOC function Thentwo palmprint ROI images can be aligned based on the dis-placement parameters and extract the mutual regions Ratiobetween the mutual region area and the area of the originalROI can be checked in the palmprint system After alignmentand common area cropping two palmprint ROI images areconstructedThenmatching distance is obtained bymatchingtwo palmprint ROI images The angular distance based onthe normalized Hamming distance is used to match the twopalmprint images for local features in palmprint biometricsystems The two palmprint images 119862

119891and 119862

119892are created

after alignment and common area cropping The matching

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Journal of Applied Mathematics 5

5 Local-Global FeatureExtraction and Matching

51 Local Feature The feature extraction technique in anypalmprint biometric system is used to obtain good interclassseparation in minimum time The features of palmprintare extracted by using 2D-DOST after the completion ofpreprocessing and ROI segmentation from palmprint Thelocal variation of instantaneous phase is used to extractfeatures from palmprint Instantaneous phase obtained usingDOST is the resolution of phase with respective to time Ithas the merits of great accurateness sturdiness to brightnessvariation and fast matching When matching two palmprintimages the angular distance based on the normalized Ham-ming distance is used

52 Global Feature DOST is used to extract the instanta-neous phase information and also it can be considered as awindowed Fourier transform The DOST can be consideredas a windowed Fourier transformThe phase-only correlation(POC) and band-limited phase-only correlation (BLPOC)are utilized to match the global information of the twopalmprint images

521 Phase-Only Correlation (POC) The Fourier transformcoefficients are considered as the global information Thetwo given Fourier transforms are matched using POC POCbased techniques have been widely used in image registrationtasks Recently a POC technique considered as a resemblancemeasure in certain biometrics systems [7 22] BLPOC ismore effective when compared to the conventional POCBLPOC is utilized to estimate the translation parametersbetween palmprint ROIs and also measure the resemblanceof the Fourier transforms of the aligned ROIs POC is anoperational method to estimate the displacement parametersbetween two palmprint images in the Fourier transformdomain Its primary principle is the translation property ofthe Fourier transforms

Consider two 1198731times 1198732images 119891(119899

1 1198992) and 119892(119899

1 1198992)

where the assumption is that the index ranges are 1198991

=

minus1198721 119872

1(1198721gt 0) and 119899

2= minus119872

2 119872

2(1198722gt 0)

For mathematical simplicity 1198731

= 21198721+ 1 and 119873

2=

21198722+ 1 Let 119865(119896

1 1198962) and 119866(119896

1 1198962) denote the 2D discrete

Fourier transforms (2D DFTs) of the two images 119865(1198961 1198962)

and 119866(1198961 1198962) are given by

119865 (1198961 1198962) = sum

11989911198992

119891 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119865(11989611198962)119890119895120579119865(11989611198962)

119866 (1198961 1198962) = sum

11989911198992

119892 (1198991 1198992)11988211989611198991

1198731

11988211989621198992

1198732

= 119860119866(11989611198962)119890119895120579119866(11989611198962)

(2)

where 1198961

= minus1198721 119872

1 1198962

= minus1198722 119872

2 1198821198731

=

119890minus119895(2120587119873

1) 1198821198732

= 119890minus119895(2120587119873

2) and the operator sum

11989911198992

denotessum1198721

1198991=minus1198721

sum1198722

1198992=minus1198722

119860119865(11989611198962)and 119860

119866(11989611198962)are amplitude com-

ponents and 119890119895120579119865(1198961 1198962) and 119890

119895120579119866(1198961 1198962) are phase compo-

nents

01

008

006

004

002

0

600

400

200

0 0200

400

600

800

6

4

2

0

0

400

200200

400

600

8

minus002

Figure 5 POC function for PolyU palmprint database

The cross spectrum 119877(1198961 1198962) between 119865(119896

1 1198962) and

119866(1198961 1198962) is given by

119877 (1198961 1198962) = 119865 (119896

1 1198962) 119866 (119896

1 1198962)

= 119860119865(11989611198962)119860119866(11989611198962)119890119895120579(11989611198962)

(3)

where 119866(1198961 1198962) denotes the complex conjugate of119866(119896

1 1198962)

and 120579(1198961 1198962) = 120579

119865(1198961 1198962) minus 120579119866(1198961 1198962) On the other hand

the cross spectrum (or normalized cross spectrum) (1198961 1198962)

is defined as

(1198961 1198962) =

119865 (1198961 1198962) 119866 (119896

1 1198962)

100381610038161003816100381610038161003816119865 (1198961 1198962) 119866 (119896

1 1198962)100381610038161003816100381610038161003816

= 119890119895120579(11989611198962)

(4)

The POC function 119903119891(1198991 1198992) = is 2D inverse discrete Fourier

transforms (IDFT) of (1198961 1198962) and is given by

119903 (1198991 1198992) =

1

1198731 1198732

(1198961 1198962)119882minus11989611198991

1198731

119882minus11989621198992

1198732

(5)

where sum11989611198962

denotes sum11987211198961=minus1198721

sum1198722

1198962=minus1198722

The most significant property of POC is compared with

the normal correlation and it gives accuracy in patternmatch-ing The distinct sharp peak is obtained when POC functionis the same The peak in the POC drops considerably if thetwo palm print images are not the same Figures 5 6 and 7show the POC function of ROI for PolyU palmprint databaseCOEP palmprint database and IITDelhi palmprint databaseThus the POC function shows much higher perceptionability than the normal correlation function In order tomatch the two palmprint images the height of the peak willbe used as a good resemblance measure

522 Band-Limited Phase-Only Correlation (BLPOC) Inthe POC based image matching method all the frequencycomponents are involved However high frequency com-ponents can be susceptible to noise BLPOC is used toeliminate noisy high frequency components BLPOC bounds

6 Journal of Applied Mathematics

01

008

006

004

002

0

minus002

600

400

200

0 0200

400600

800

0

400

200

0 0200

400600

8

Figure 6 POC function for COEP palmprint database

006

004

002

0

minus002

600

400

200

0 0

200

400

600800

4

2

0

2

0

400

200 400

6008

Figure 7 POC function for IIT Delhi palmprint database

the range of spectrum of the given palmprint image BLPOCfunction between two ROI images can be measured as thePOC function between their low-pass filtered versions basedon the definition of BLPOC Therefore BLPOC functioncan preserve the properties of the POC function BLPOCfunction got a distinct peak if two ROI images are thesame But the locality of peak in two ROI images givesthe translational displacement Experimentations specify thatBLPOC function provides a much higher perception abilitythan the original POC function in palmprint biometricsystem Therefore the efficient size of spectrum is given by1198711= 21198800+ 1 and 119871

2= 21198810+ 1 BLPOC function is given by

11990111988001198810

119892119891(119898 119899) =

1

11987111198712

1198800

sum

119906=minus1198800

1198810

sum

V=minus1198810

119877119866119865(119906V)1198901198952120587(1198981199061198711+119899V1198712) (6)

where 119898 = minus1198800 119880

0 and 119899 = minus119881

0 119881

0 where 119906 =

minus1198800 119880

0 and V = minus119881

0 119881

0 where 0 le 119880

0le 1198720

0 le 1198810le 1198730

Figures 8 9 and 10 are BLPOC function for PolyUpalmprint database COEP palmprint database and IITDelhipalmprint database respectively These examples indicatethat in the instance of genuine matching (a matching per-formed between a pair of palmprint images from the samehand) the BLPOC will show a much sharper peak than POC

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 8 BLPOC function for PolyU palmprint database

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 9 BLPOC function for COEP palmprint database

015

01

005

0

minus005

300

200

100

0 0100

200

300400200

100 200

300

Figure 10 BLPOC function for IIT Delhi palmprint database

however for an imposter matching (a matching performedbetween a couple of palmprint images from different hands)neither BLPOC nor POC will display a different severe peakBLPOC function can be adopted to align the translationbetween two palmprint ROI images and then to measure theresemblance between Fourier transforms of the aligned ROIs

Journal of Applied Mathematics 7

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 11 ROC curve using local features of proposed DOST forPolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 12 ROC curve using global features of proposed DOST forPolyU palmprint database

6 Local-Global Information Fusion (LGIF) forPalmprint Recognition

LGIF based palmprint recognition algorithm is described inthis section The palmprint image acquisition device and theROI extraction systemcandecrease the symmetrical transfor-mations between intraclass ROIs But it is quiet unavoidablebecause there is some translation between intraclass ROIsin the palmprint images These will deteriorate the genuinematching scores This problem is solved by transformingthe given information in horizontal and vertical directionsnumerous times and the smallest of the subsequent similardistances is measured as the final matching distance Thisproblem can be solved in a dissimilar way by calculating thedisplacement parameters between two ROIs using BLPOC

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 13 ROC curve using local features of proposed DOST forCOEP palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 14 ROC curve using global features of proposed DOST forCOEP palmprint database

function Then the mutual regions can be cropped based onwhich the pattern matching is completed The displacementparameters between two palmprint ROI images can be calcu-lated from the peak locality of global BLPOC function Thentwo palmprint ROI images can be aligned based on the dis-placement parameters and extract the mutual regions Ratiobetween the mutual region area and the area of the originalROI can be checked in the palmprint system After alignmentand common area cropping two palmprint ROI images areconstructedThenmatching distance is obtained bymatchingtwo palmprint ROI images The angular distance based onthe normalized Hamming distance is used to match the twopalmprint images for local features in palmprint biometricsystems The two palmprint images 119862

119891and 119862

119892are created

after alignment and common area cropping The matching

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

6 Journal of Applied Mathematics

01

008

006

004

002

0

minus002

600

400

200

0 0200

400600

800

0

400

200

0 0200

400600

8

Figure 6 POC function for COEP palmprint database

006

004

002

0

minus002

600

400

200

0 0

200

400

600800

4

2

0

2

0

400

200 400

6008

Figure 7 POC function for IIT Delhi palmprint database

the range of spectrum of the given palmprint image BLPOCfunction between two ROI images can be measured as thePOC function between their low-pass filtered versions basedon the definition of BLPOC Therefore BLPOC functioncan preserve the properties of the POC function BLPOCfunction got a distinct peak if two ROI images are thesame But the locality of peak in two ROI images givesthe translational displacement Experimentations specify thatBLPOC function provides a much higher perception abilitythan the original POC function in palmprint biometricsystem Therefore the efficient size of spectrum is given by1198711= 21198800+ 1 and 119871

2= 21198810+ 1 BLPOC function is given by

11990111988001198810

119892119891(119898 119899) =

1

11987111198712

1198800

sum

119906=minus1198800

1198810

sum

V=minus1198810

119877119866119865(119906V)1198901198952120587(1198981199061198711+119899V1198712) (6)

where 119898 = minus1198800 119880

0 and 119899 = minus119881

0 119881

0 where 119906 =

minus1198800 119880

0 and V = minus119881

0 119881

0 where 0 le 119880

0le 1198720

0 le 1198810le 1198730

Figures 8 9 and 10 are BLPOC function for PolyUpalmprint database COEP palmprint database and IITDelhipalmprint database respectively These examples indicatethat in the instance of genuine matching (a matching per-formed between a pair of palmprint images from the samehand) the BLPOC will show a much sharper peak than POC

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 8 BLPOC function for PolyU palmprint database

015

01

005

0

minus005300

200

100

0

300

200

1000

400200

100

300

200

100

4

Figure 9 BLPOC function for COEP palmprint database

015

01

005

0

minus005

300

200

100

0 0100

200

300400200

100 200

300

Figure 10 BLPOC function for IIT Delhi palmprint database

however for an imposter matching (a matching performedbetween a couple of palmprint images from different hands)neither BLPOC nor POC will display a different severe peakBLPOC function can be adopted to align the translationbetween two palmprint ROI images and then to measure theresemblance between Fourier transforms of the aligned ROIs

Journal of Applied Mathematics 7

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 11 ROC curve using local features of proposed DOST forPolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 12 ROC curve using global features of proposed DOST forPolyU palmprint database

6 Local-Global Information Fusion (LGIF) forPalmprint Recognition

LGIF based palmprint recognition algorithm is described inthis section The palmprint image acquisition device and theROI extraction systemcandecrease the symmetrical transfor-mations between intraclass ROIs But it is quiet unavoidablebecause there is some translation between intraclass ROIsin the palmprint images These will deteriorate the genuinematching scores This problem is solved by transformingthe given information in horizontal and vertical directionsnumerous times and the smallest of the subsequent similardistances is measured as the final matching distance Thisproblem can be solved in a dissimilar way by calculating thedisplacement parameters between two ROIs using BLPOC

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 13 ROC curve using local features of proposed DOST forCOEP palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 14 ROC curve using global features of proposed DOST forCOEP palmprint database

function Then the mutual regions can be cropped based onwhich the pattern matching is completed The displacementparameters between two palmprint ROI images can be calcu-lated from the peak locality of global BLPOC function Thentwo palmprint ROI images can be aligned based on the dis-placement parameters and extract the mutual regions Ratiobetween the mutual region area and the area of the originalROI can be checked in the palmprint system After alignmentand common area cropping two palmprint ROI images areconstructedThenmatching distance is obtained bymatchingtwo palmprint ROI images The angular distance based onthe normalized Hamming distance is used to match the twopalmprint images for local features in palmprint biometricsystems The two palmprint images 119862

119891and 119862

119892are created

after alignment and common area cropping The matching

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Journal of Applied Mathematics 7

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 11 ROC curve using local features of proposed DOST forPolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 12 ROC curve using global features of proposed DOST forPolyU palmprint database

6 Local-Global Information Fusion (LGIF) forPalmprint Recognition

LGIF based palmprint recognition algorithm is described inthis section The palmprint image acquisition device and theROI extraction systemcandecrease the symmetrical transfor-mations between intraclass ROIs But it is quiet unavoidablebecause there is some translation between intraclass ROIsin the palmprint images These will deteriorate the genuinematching scores This problem is solved by transformingthe given information in horizontal and vertical directionsnumerous times and the smallest of the subsequent similardistances is measured as the final matching distance Thisproblem can be solved in a dissimilar way by calculating thedisplacement parameters between two ROIs using BLPOC

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 13 ROC curve using local features of proposed DOST forCOEP palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 14 ROC curve using global features of proposed DOST forCOEP palmprint database

function Then the mutual regions can be cropped based onwhich the pattern matching is completed The displacementparameters between two palmprint ROI images can be calcu-lated from the peak locality of global BLPOC function Thentwo palmprint ROI images can be aligned based on the dis-placement parameters and extract the mutual regions Ratiobetween the mutual region area and the area of the originalROI can be checked in the palmprint system After alignmentand common area cropping two palmprint ROI images areconstructedThenmatching distance is obtained bymatchingtwo palmprint ROI images The angular distance based onthe normalized Hamming distance is used to match the twopalmprint images for local features in palmprint biometricsystems The two palmprint images 119862

119891and 119862

119892are created

after alignment and common area cropping The matching

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

8 Journal of Applied Mathematics

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 15 ROC curve using local features of proposed DOST forIIT Delhi palmprint database

0

1

Proposed workOrdinal codePalm code

09

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Figure 16 ROC curve using global features of proposed DOST forIIT Delhi palmprint database

distance119889119871is obtained bymatching the two palmprint images

119862119891and 119862

119892 The peak value of the BLPOC function between

two palmprint images is used to measure the similarity oftheir Fourier transformsThe peak value is indicated as pocStherefore the matching distance is termed as 119889

119866= 1 minus pocS

The two matching distances 119889119871and 119889

119866are obtained We

got the final matching distance by fusing the two distances119889119871and 119889

119866 In BLPOC function the peak value between

palmprint ROI images is used to calculate the resemblanceof their Fourier transforms to obtain matching distance Inthis paper two matching distances are obtained from twodissimilar matchers namely local feature based matcher(matcher 1) and global feature based matcher (matcher 2)and the maximum weighted (MW) rule can be adoptedThe weights are assigned according to the equal error rate

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 17 ROC curve using both local and global features ofproposed DOST for PolyU palmprint database

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 18 ROC curve using both local and global features ofproposed DOST for COEP palmprint database

(EER) obtained on a training dataset by different matchersbased on MW fusion rule It is obvious that the weights areinversely proportional to the corresponding EERs Then thefinal matching distance is estimated

7 Experimental Results and Discussion

Three different sets of hand image databases are used for theproposed system The PolyU palmprint database [27] con-tains 7752 grayscale images corresponding to 386 differentpalms inBMP image formatThe twenty samples are collectedin two different sessions 10 samples were collected from firstsession and 10 samples were captured in the second sessionrespectively

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Journal of Applied Mathematics 9

0

109

08

07

06

05

04

03

02

01

100

10minus1

10minus2

10minus3

10minus4

10minus5

Gen

uine

acce

ptan

ce ra

te (

)

False acceptance rate ()

Proposed workOrdinal codePalm code

Figure 19 ROC curve using both local and global features ofproposed DOST for IIT Delhi palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 20Distance distributions of genuinematching and impostermatching of proposed DOST for PolyU palmprint database

COEP palmprint database contains [28] 1344 imagescorresponding to 168 people Eight different images arecollected from single personrsquos palm

IIT Delhi palmprint database [29] contains 230 subjectsThe subjects in the palmprint database are collected fromthe age group 12ndash57 years Seven images per each subjectare collected from the left and right hand An identification

Table 1 Performance comparison of different palmprint verificationschemes for both local and global combination

CRR () EER () DI(a) Performance for PolyU database

Palm code [9] 9991 05338 55807Ordinal code [21] 10000 00709 66785Proposed 10000 00035 72541

(b) Performance for COEP databasePalm code [9] 9961 36732 30734Ordinal code [21] 9984 17544 33966Proposed 10000 12605 40228

(c) Performance for IIT Delhi databasePalm code [9] 9968 52108 1800Ordinal code [21] 10000 11762 22174Proposed 10000 10217 30168

Table 2 Computation time for key processes both local and globalcombination

Extraction (ms) Matching (ms) Total (ms)(a) Systems speed for PolyU database

Proposed 5456 15 5606(b) Systems speed for COEP database

Proposed 8541 307 8848(c) Systems speed for IIT Delhi database

Proposed 8334 315 8649

number is assigned to every user for the acquired imagesThe resolution of these images is 800 times 600 pixels Afterpreprocessing original images 150 times 150 pixels are automat-ically cropped and considered as ROI

Verification means by evaluating one or more individualgenetic traits of a person can be uniquely identified In thispaper all the classes of palmprint images were involved Themetrics such as EER correct recognition rate (CRR) andreceiver operating characteristics (ROC) are used to measurethe recognition accuracyTheCRR of a system can be definedby

CRR =1198771

1198772times 100 (7)

where1198771 denotes the number of correct recognition of palm-print images and 1198772 is the total number of palmprint imagesPerformance comparison of different palmprint verificationschemes is shown in Table 1 Table 2 shows the computationtime for key processes

ROC curves are used to analyse the performance ofthe palmprint biometric system ROC is a plot of genuineacceptance rate (GAR) against false acceptance rate (FAR)and shown in Figures 11 12 13 14 15 16 17 18 and 19for PolyU palmprint database COEP palmprint databaseand IIT Delhi palmprint database The FAR and FRR of apalmprint biometric system can be obtained by adjusting thevarious thresholds It is known that both FAR and FRR are

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

10 Journal of Applied Mathematics

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 21 Distance distributions of genuinematching and impostermatching of proposed DOST for COEP palmprint database

35

3

25

2

15

1

05

0

01 02 03 04 05

Matching distance

ImposterGenuine

()

Figure 22Distance distributions of genuinematching and impostermatching of proposed DOST for IIT Delhi palmprint database

inversely related Thus at a particular threshold these twocurves intersect This intersect point is termed as an EERThe EER is defined at the FAR which is equal to the FRRThe distance distributions of genuine matching and impostermatching acquired by proposed scheme are plotted in Figures20 21 and 22 The decidability index (DI) is defined as how

well the genuine and the imposter distributions are separatedwhich is defined as

DI =10038161003816100381610038161205831 minus 1205832

1003816100381610038161003816

radic(1205902

1+ 1205902

2) 2

(8)

where 1205831(1205832)is the mean of the genuine (imposter) matching

distances and 1205901(1205902) is the standard deviation of the genuine

(imposter) matching distances Then performance of anybiometric system can bemeasured by its1198891015840 score among othermetrics

8 Conclusion

In this work palmprint images are used for recognition LGIFbased palmprint features are critical for the image recognitionand observation They play dissimilar and correspondingroles in a system In LGIF the local instantaneous phaseinformation obtained by the DOST technique is measuredas the local feature From the perception of time-frequencyanalysis if the scale of the DOST goes to infinity it convertsto the Fourier transformTheglobal information of palmprintrecognition system refines the alignment of palmprint imagein pattern matching Extensive investigational outcomesshown on the palmprint database indicate that the proposedsystem could achieve much better performance in terms ofEER

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

The authors would like to thank Hong Kong PolytechnicUniversity IIT-Delhi and College of Engineering-Pune forproviding the palmprint databases

References

[1] A K Jain P J Flynn and A Ross Handbook of BiometricsSpringer Berlin Germany 2010

[2] D Maltoni D Maio A K Jain and S Prabhakar Handbook ofFingerprint Recognition Springer London UK 2009

[3] N Ratha and R Bolle Automatic Fingerprint RecognitionSystems Springer New York NY USA 2011

[4] R Belguechi E Cherrier C Rosenberger and S Ait-AoudialdquoOperational bio-hash to preserve privacy of fingerprint minu-tiae templatesrdquo IET Biometrics vol 2 no 2 pp 76ndash84 2013

[5] J Gu J Zhou and C Yang ldquoFingerprint recognition bycombining global structure and local cuesrdquo IEEE Transactionson Image Processing vol 15 no 7 pp 1952ndash1964 2006

[6] F Yuchun T Tieniu and W Yunhong ldquoFusion of globaland local features for face verificationrdquo in Proceedings of theInternational Conference on Pattern Recognition pp 382ndash385Quebec Canada August 2002

[7] K Miyazawa K Ito T Aoki K Kobayashi and H NakajimaldquoAn effective approach for Iris recognition using phase-based

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Journal of Applied Mathematics 11

image matchingrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 30 no 10 pp 1741ndash1756 2008

[8] Z Sun Y Wang T Tan and J Cui ldquoImproving iris recognitionaccuracy via cascaded classifiersrdquo IEEETransactions on SystemsMan and Cybernetics C Applications and Reviews vol 35 no 3pp 435ndash441 2005

[9] D Zhang W-K Kong J You and M Wong ldquoOnline palm-print identificationrdquo IEEE Transactions on Pattern Analysis andMachine Intelligence vol 25 no 9 pp 1041ndash1050 2003

[10] C C Han H L Cheng C L Lin and K C Fan ldquoPersonalauthentication using palm-print featuresrdquo Pattern Recognitionvol 36 no 2 pp 371ndash381 2003

[11] H Matos P Oliveira and F Magalhaes ldquoHand-geometry basedrecognition systemrdquo in Image Analysis and Recognition vol7325 of Lecture Notes in Computer Science pp 38ndash45 Springer2012

[12] D Huang Y Tang Y Wang L Chen and Y Wang ldquoHand veinrecognition based on oriented gradient maps and local featurematchingrdquo in Computer Vision vol 7727 of Lecture Notes inComputer Science pp 430ndash444 Springer New York NY USA2013

[13] L Nanni and A Lumini ldquoA multi-matcher system based onknuckle-based featuresrdquo Neural Computing and Applicationsvol 18 no 1 pp 87ndash91 2009

[14] A Kumar and Y Zhou ldquoPersonal identification using fingerknuckle orientation featuresrdquo Electronics Letters vol 45 no 20pp 1023ndash1025 2009

[15] H Hollein Forensic Voice Identification Academic Press 2002[16] C Middendorff Multi-Biometric Approaches to Ear Biometrics

and Soft Biometrics BiblioBazaar 2011[17] C McCool R Wallace M McLaren L El Shafey and S

Marcel ldquoSession variability modelling for face authenticationrdquoIET Biometrics vol 2 no 3 pp 117ndash129 2013

[18] T H Nguyen Y Wang Y Ha and R Li ldquoImproved chaffpoint generation for vault scheme in bio-cryptosystemsrdquo IETBiometrics vol 2 no 2 pp 48ndash55 2013

[19] N Duta A K Jain and K VMardia ldquoMatching of palmprintsrdquoPattern Recognition Letters vol 23 no 4 pp 477ndash485 2002

[20] J Funada N Ohta M Mizoguchi et al ldquoFeature extractionmethod for palmprint considering elimination of creasesrdquo inProceedings of the International Conference on Pattern Recogni-tion pp 1849ndash1854 Brisbane Australia August 1998

[21] Z N Sun T Tan Y Wang and S Z Li ldquoOrdinal palmprintrepresention for personal identification [represention readrepresentation]rdquo in Proceedings of the IEEE Computer SocietyConference on Computer Vision and Pattern Recognition (CVPRrsquo05) vol 1 pp 279ndash284 June 2005

[22] Y Su S Shan X Chen and W Gao ldquoHierarchical ensemble ofglobal and local classifiers for face recognitionrdquo IEEE Transac-tions on Image Processing vol 18 no 8 pp 1885ndash1896 2009

[23] F Li K H Leung and X Yu ldquoA two-level matching schemefor speedy and accurate palmprint identificationrdquo in Advancesin Multimedia Modeling vol 4352 of Lecture Notes in ComputerScience pp 323ndash332 Springer Berlin Germany 2006

[24] X Pan Q Ruan and Y Wang ldquoPalmprint recognition usingfusion of local and global featuresrdquo in Proceedings of theInternational Conference on Intelligence Signal Processing andCommunication System pp 642ndash645 Xiamen China Novem-ber 2007

[25] S B Nikam P Goel R Tapadar and S Agarwal ldquoCombininggabor local texture pattern and wavelet global features for fin-gerprint matchingrdquo in Proceedings of the International Confer-ence on Computational Intelligence andMultimedia Applications(ICCIMA rsquo07) pp 409ndash416 Washington DC USA 2007

[26] G S Badrinath and P Gupta ldquoStockwell transform based palm-print recognitionrdquo Applied Soft Computing vol 11 no 7 pp4267ndash4281 2011

[27] ldquoPolyU Palmprint Databaserdquo The Hong Kong PolytechnicUniversity July 2010 httpwww4comppolyueduhksimbio-metrics

[28] COEP Palm Print Database (College of Engineering Pune)httpwwwcoeporgin

[29] ldquoPalmprint Image Database version 1rdquo IIT Delhi India 2007httpwww4comppolyueduhksimcsajaykrIITDDatabasePalmhtm

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of

Submit your manuscripts athttpwwwhindawicom

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical Problems in Engineering

Hindawi Publishing Corporationhttpwwwhindawicom

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

OptimizationJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Operations ResearchAdvances in

Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

International Journal of Mathematics and Mathematical Sciences

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Algebra

Discrete Dynamics in Nature and Society

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Stochastic AnalysisInternational Journal of