cs484 color
TRANSCRIPT
-
8/19/2019 Cs484 Color
1/27
Color Image Processing
Selim Aksoy
Department of Computer Engineering
Bilkent University
-
8/19/2019 Cs484 Color
2/27
CS !"!# Spring $%&& '$%& Selim Aksoy $
Color
Use (eavily in (uman vision. )isible spectrum for (umans is !%% nm
*blue+ to ,%% nm *re+.
-ac(ines can see/ muc( more0 e.g.# 12rays# infrare# raio 3aves.
Aapte from 4on5ales an 6oos
-
8/19/2019 Cs484 Color
3/27
CS !"!# Spring $%&& '$%& Selim Aksoy 7
8uman visual system
Color perception 9ig(t (its t(e retina# 3(ic( contains p(otosensitivecells.
:(ese cells convert t(e spectrum into a fe3 iscretevalues.
Aapte from Steve Seit5# U of6as(ington
-
8/19/2019 Cs484 Color
4/27
CS !"!# Spring $%&& '$%& Selim Aksoy !
8uman visual system
:(ere are t3o types of p(otosensitive cells; Cones
Sensitive to colore lig(t# but not very sensitive to im lig(t.
!> nm *4"% nm *
-
8/19/2019 Cs484 Color
5/27
CS !"!# Spring $%&& '$%& Selim Aksoy >
actors t(at a=ect perception
9ig(t; t(e spectrum of energy t(atilluminates t(e obect surface.
-
8/19/2019 Cs484 Color
6/27
CS !"!# Spring $%&& '$%& Selim Aksoy
Color moels
:(ey provie a stanar 3ay of specifying aparticular color using a 7D coorinatesystem.
8ar3are oriente
-
8/19/2019 Cs484 Color
7/27CS !"!# Spring $%&& '$%& Selim Aksoy ,
Aitive an subtractive colors
Aapte from 4on5ales an 6oos
-
8/19/2019 Cs484 Color
8/27CS !"!# Spring $%&& '$%& Selim Aksoy "
-
8/19/2019 Cs484 Color
9/27CS !"!# Spring $%&& '$%& Selim Aksoy G
C- moel
Cyan2-agenta2ello3 is a subtractive moel3(ic( is goo to moel absorption of colors.
Appropriate for paper printing.
−
=
B
G
R
Y
M
C
1
1
1
Aapte from ctavia Camps# Penn State
-
8/19/2019 Cs484 Color
10/27
CS !"!# Spring $%&& '$%& Selim Aksoy &%
CIE c(romaticity moel
:(e Commission Internationale elHEclairage ene 7 stanar primaries; 1#
# J t(at can be ae to form all visiblecolors.
3as c(osen so t(at its color matc(ingfunction matc(es t(e sum of t(e 7 (umancone responses.
−
−−
−−
=
Z
Y
X
B
G
R
8986.01185.00583.0
0283.09984.19843.0
2883.05326.09107.1
=
B
G
R
Z
Y
X
1149.10661.00000.0
1143.05868.02988.0
2001.01736.06067.0
Aapte from ctavia Camps# Penn State
-
8/19/2019 Cs484 Color
11/27
CS !"!# Spring $%&& '$%& Selim Aksoy &&
CIE c(romaticity moel
K# y# 5 normali5e 1# # Jsuc( t(at
K L y L 5 M &.
Actually only K an y
are neee because5 M & 2 K 2 y.
Pure colors are at t(ecurve bounary.
6(ite is *&N7# &N7# &N7+.
Aapte from ctavia Camps# Penn State
-
8/19/2019 Cs484 Color
12/27
CS !"!# Spring $%&& '$%& Selim Aksoy &$
CIE 9ab *9OaOb+ moel
ne luminance c(annel *9+an t3o color c(annels *a anb+.
In t(is moel# t(e colori=erences 3(ic( you
perceive correspon toEucliian istances in CIE 9ab.
:(e a aKis eKtens from green*2a+ to re *La+ an t(e b aKis
from blue *2b+ to yello3 *Lb+. :(e brig(tness *9+ increasesfrom t(e bottom to t(e top oft(e 7D moel.
(ttp;NN333.f(o2emen.eN(o=mannNcielab%7%$$%%7.pf Aapte from 9ina S(apiro# U of
6as(ington
-
8/19/2019 Cs484 Color
13/27
CS !"!# Spring $%&& '$%& Selim Aksoy &7
IF moel
8ave better compression properties. 9uminance is encoe using more bits
t(an c(rominance values I an F *(umansare more sensitive to t(an I an F+.
9uminance use by blackN3(ite :)s.
All 7 values use by color :)s.
−
−−=
B
G R
Q
I Y
311.0532.0212.0
321.0275.0596.0114.0587.0299.0
Aapte from ctavia Camps# Penn State
-
8/19/2019 Cs484 Color
14/27
CS !"!# Spring $%&& '$%& Selim Aksoy &!
8S) moel
8S); 8ue# saturation# value are non2linearfunctions of
−−+−
−+−−=
-
8/19/2019 Cs484 Color
15/27
CS !"!# Spring $%&& '$%& Selim Aksoy &>
8S) moel
Uniform; eQual *small+ steps give t(e sameperceive color c(anges.
8ue is encoe as an angle *% to $π+. Saturation is t(e istance to t(e vertical aKis *% to
&+. Intensity is t(e (eig(t along t(e vertical aKis *% to
&+.
Aapte from 4on5ales an 6oos
-
8/19/2019 Cs484 Color
16/27
CS !"!# Spring $%&& '$%& Selim Aksoy &
8S) moel
*9eft+ Image of foo originating from a igital camera.
*Center+ Saturation value of eac( piKel ecrease $%R.
*
-
8/19/2019 Cs484 Color
17/27
CS !"!# Spring $%&& '$%& Selim Aksoy &,
Color moels
C-
-
8/19/2019 Cs484 Color
18/27
CS !"!# Spring $%&& '$%& Selim Aksoy &"
EKamples; pseuocolor
-
8/19/2019 Cs484 Color
19/27
CS !"!# Spring $%&& '$%& Selim Aksoy &G
EKamples; pseuocolor
-
8/19/2019 Cs484 Color
20/27
CS !"!# Spring $%&& '$%& Selim Aksoy $%
EKamples; pseuocolor
-
8/19/2019 Cs484 Color
21/27
CS !"!# Spring $%&& '$%& Selim Aksoy $&
EKamples; segmentation
Can cluster on color values an piKellocations.
Can use connecte components an anapproKimate color criteria to n regions.
Can train an algorit(m to look for certaincolore regions T for eKample# skin color.
riginal
-
8/19/2019 Cs484 Color
22/27
CS !"!# Spring $%&& '$%& Selim Aksoy $$
EKamples; segmentation
Skin color in
-
8/19/2019 Cs484 Color
23/27
CS !"!# Spring $%&& '$%& Selim Aksoy $7
EKamples; segmentation
*9eft+ Input vieo frame.
*Center+ PiKels classie accoring to
-
8/19/2019 Cs484 Color
24/27
CS !"!# Spring $%&& '$%& Selim Aksoy $!
EKamples; (istogram
8istogram is fast an easy to compute. Si5e can easily be normali5e so t(at
i=erent image (istograms can becompare.
Can matc( color (istograms for atabaseQuery or classication.
Aapte from 9ina S(apiro# U of6as(ington
-
8/19/2019 Cs484 Color
25/27
CS !"!# Spring $%&& '$%& Selim Aksoy $>
EKamples; (istogram
-
8/19/2019 Cs484 Color
26/27
CS !"!# Spring $%&& '$%& Selim Aksoy $
EKamples; image retrieval
Aapte from 9ina S(apiro# U of6as(ington
-
8/19/2019 Cs484 Color
27/27
CS !"! Spring $%&& '$%&& Selim Aksoy $,
Summary
:o print *