acfrogck1rp9i3e_ud3nlgyzs806gunykjmxzrba8npnskn6x6g8yuftfart4-1adaa2ezkrd01iuhu7vhdtrg6fklwpksauaa89gv7wruqmbqa2xocm1noo4zl_zhe=...

26
25 December 2014 1 Department of IT WAVELET- BASED APPROACH TO CHARACTER SKELETON MEMBERS: GUIDE:

Upload: sooriya7

Post on 18-Jul-2016

213 views

Category:

Documents


0 download

DESCRIPTION

vcvc

TRANSCRIPT

25 December 20141

Department of IT

WAVELET- BASED APPROACH TO CHARACTER SKELETON

MEMBERS:

GUIDE:

25 December 20142

Department of IT

PROBLEM STATEMENT

To implement a character skeletonization scheme which is effective in extracting relevant features of the character for optical character recognition(OCR), handwriting recognition, signature verification.

25 December 20143

Department of IT

ANALYSIS PHASE

25 December 20144

Department of IT

LITERATURE SURVEY

• Wavelet-based approach to character skeleton.

• Skeletonization of ribbon-like shapes based on new wavelet function.

25 December 20145

Department of IT

PROBLEMS TO OVERCOME

In traditional algorithms (Fourier Transform-FT and Symmetric axis transform-SAT),

• The symmetric center of skeleton was computed indirectly by measuring inscribed circle or the symmetric triangle.

• Moreover, the skeleton obtained slightly deviates from the center.

• Skeleton obtained is distorted by artifacts and branches.

• Cannot be applied to gray-level images

25 December 20146

WHAT IS CHARACTER SKELETON?

• The ‘skeleton’ of a character is the locus of the midpoints or the symmetric axis of the character stroke

25 December 20147

WHY WAVELET APPROACH?

• Wavelet transform is capable of providing the time and frequency information simultaneously.

• Modulus-maxima symmetric analysis technique to directly extract skeleton by contour line detection.

• Skeleton can be extracted more accurately based on wavelet based amendment processing.

• Robust against noise.

• Applicable to both binary and gray-level images.

25 December 20148

Department of IT

PROPOSED MODELIn our new wavelet-based approach ,these are the ideas

proposed to overcome the above problems,• The symmetric center can be determined as the midpoints

of the symmetric points directly using the new wavelet-based symmetry analysis method.

• A skeleton representation is robust against noise and insensitive to linear geometric transformations, such as translation, rotation, and scaling.

• The Amendment processing method of the proposed model plays an important role in effectively removing artifacts and distortions.

• Unlike the traditional pixel-based methods the proposed curve-based method is applicable to gray-level images too.

25 December 20149

Department of IT

USE-CASE DIAGRAM

Detect contour points and join them

user character recognition systemReduce noise by thresholding

Compute modulus of WT and gradient codes

Perform wavelet transform

Input the digitized character image

Amendment processing

25 December 201410

Department of IT

SYSTEM TEST PLAN

INPUTEnglish characters &numerical images

OUTPUTSkeleton of the Character stroke.

25 December 201411

Department of IT

DESIGN PHASE

25 December 201412

Department of IT

MODULES

There are two major modules, namely

1. Generating the new wavelet function, define the low-pass filter and wavelets.

2. Using the WLMMS technique, extracting the skeleton of the character stroke.

25 December 201413

Department of IT

EXTRACTION OF SKELETONIn this module,• A direct technique is used, where a new wavelet-based

symmetry analysis is developed to find the central line of the stroke.

25 December 201414

Department of IT

WAVELETS

• The wavelets imply the partial derivatives of a low pass filter

and

is chosen as

),(),(1 yxx

yx

),(),(2 yxy

yx

),( yx )( 22 yx

25 December 201415

ALGORITHM

• Input the digitized character image and select a suitable scale.

• Perform the wavelet transform to the character image.

• Compute modulus of the wavelet transform.• Remove the noise.• Compute local modulus maxima .• Find the symmetric pair of modulus maxima

25 December 201416

Department of IT

AMENDMENT PROCESSING

Here, the obtained skeleton of the character stroke is modified to remove spurs and artifacts so that the skeleton is accurate and relevant for recognition.

25 December 201417

Department of IT

UNIT TEST PLAN

• Unit 1- Primary Skeleton Extraction Unit In which a character image is inputted and the

primary skeleton of the regular region is obtained as output.

• Unit 2- Amendment Processing UnitIn which output of the unit 1 is used as input

and modified to obtain the final skeleton of the input image as output.

25 December 201418

Department of IT

DESIGN• The wavelet transform of the image is performed using the function

• The modulus of the WT is found using

and gradient direction with the help of

• For the noise removal, the threshold level chosen is 0.39.

2221 ),(),(|:),(| yxfWyxfWyxfW ssS

)),(),(tan(arg),( 1

2

yxfWyxfWyxfA

s

ss

),)(*(),)(*(),( 11 yxfx

syxfyxfW sss

),)(*(),)(*(),( 22 yxfy

syxfyxfW sss

25 December 201419

Department of IT

SOFTWARE AND HARDWARE

• SOFTWARE:Mat lab 7.0

• HARDWARE:3.0 GHz Pentium IV CPU1GB RAM

25 December 201420

Department of IT

SCREEN SHOTS

25 December 201421

Department of IT

ORIGINAL IMAGE

25 December 201422

Department of IT

GRADIENT DIRECTION IMAGE

25 December 201423

Department of IT

MODULUS MAXIMA IMAGE

25 December 201424

Department of IT

PRIMARY SKELETON

25 December 201425

Department of IT

FINAL SKELETON

25 December 201426

THANK YOU

Department of IT