colorppt
TRANSCRIPT
-
7/24/2019 colorppt
1/53
Digital Image ProcessingChapter 5:
Color Image Processing
-
7/24/2019 colorppt
2/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Spectrum of White LightSpectrum of White Light
1666 ir Isaac !e"ton, 2# $ear old, disco%ered "hite light s&ectr'm.
-
7/24/2019 colorppt
3/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Electromagnetic SpectrumElectromagnetic Spectrum
isi)le light "a%elength* from aro'nd #++ to ++ nm
1. -or an achromatic (monochrome light so'rce,
there is onl$ 1 attri)'te to descri)e the /'alit$*intensit$
2. -or a chromatic light so'rce, there are 0 attri)'tes to descri)ethe /'alit$*
Radiance total amo'nt of energ$ flo" from a light so'rce (Watts
'minance amo'nt of energ$ recei%ed )$ an o)ser%er (l'mens
3rightness intensit$
-
7/24/2019 colorppt
4/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Sensitivity of Cones in the Human EyeSensitivity of Cones in the Human Eye
64 millions cones
in a h'man e$e 4 65 sensiti%e to Red light
4 00 sensiti%e to Green light
4 2 sensiti%e to 3l'e light
Primar$ colors*
Defined CIE in 1701
Red ++ nm
Green 5#6.1nm
3l'e #05.8 nm
CIE Commission Internationale de l9Eclairage
(:he International Commission on Ill'mination
-
7/24/2019 colorppt
5/53
Primary and Secondary ColorsPrimary and Secondary Colors
Primar$color
Primar$
color
Primar$
color
econdar$
colors
-
7/24/2019 colorppt
6/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Primary and Secondary Colors (cont.)Primary and Secondary Colors (cont.)
;dditi%e &rimar$ colors* RG3
'se in the case of light so'rces
s'ch as color monitors
')tracti%e &rimar$ colors* C
-
7/24/2019 colorppt
7/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
R! Color "odelR! Color "odel
P'r&ose of color models* to facilitate the s&ecification of colors in
some standard
RG3 color models*4)ased on Cartesian
coordinate s$stem.
:he n'm)er of )its 'sed
to re&resent each &i?el in
RG3 &ace is called the
&i?el De&th.
-
7/24/2019 colorppt
8/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
R! Color Cu#eR! Color Cu#e
R 8 )its
G 8 )its3 8 )its
Color de&th 2# )its
16216 colors
@idden faces
of the c')e
-
7/24/2019 colorppt
9/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
R! Color "odel (cont.)R! Color "odel (cont.)
Red fi?ed at 12
-
7/24/2019 colorppt
10/53
C"$ and C"$% Color "odelsC"$ and C"$% Color "odels
=
B
G
R
Y
M
C
1
1
1
C C$an
<
-
7/24/2019 colorppt
11/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
HS& Color "odelHS& Color "odel
RG3, C
-
7/24/2019 colorppt
12/53
@'e* dominant color corres&onding to a dominant
"a%elength of mi?t're light "a%e.(dominant color as
&ercei%ed )$ an o)ser%er.
at'ration* Relati%e &'rit$ or amo'nt of "hite light mi?ed
"ith a h'e (in%ersel$ &ro&ortional to amo'nt of "hite
light added
3rightness* Intensit$
Color Characteri'ationColor Characteri'ation
@'e
at'rationChromaticit$
amo'nt of red (B, green (= and )l'e ( to form an$ &artic'lar
color is called tristimulus.
-
7/24/2019 colorppt
13/53
Hue and Saturation on Color PlanesHue and Saturation on Color Planes
1. ; dot is the &lane is an ar)itrar$ color
2. @'e is an angle from a red a?is.
0. at'ration is a distance to the &oint.
-
7/24/2019 colorppt
14/53
HS& Color "odel (cont.)HS& Color "odel (cont.)
Intensit$ is gi%en )$ a &osition on the %ertical a?is.
-
7/24/2019 colorppt
15/53
HS& Color "odelHS& Color "odel
Intensit$ is gi%en )$ a &osition on the %ertical a?is.
-
7/24/2019 colorppt
16/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Eample HS& Components of R! Cu#eEample HS& Components of R! Cu#e
@'e at'ration Intensit$
RG3 C')e
-
7/24/2019 colorppt
17/53
Converting Colors from R! to HS&Converting Colors from R! to HS&
>
=
GB
GBH
if06+
if
[ ]
[ ]
+
+=
212
1
(((
((2
1
cosBGBRGR
BRGR
BGRS
++=
01
.(01 BGRI ++=
-
7/24/2019 colorppt
18/53
Converting Colors from HS& to R!Converting Colors from HS& to R!
.1( SIB =
+=
6+cos(cos1HHSIR
.(1 BRG +=
RG sector* 12++
-
7/24/2019 colorppt
19/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Eample HS& Components of R! ColorsEample HS& Components of R! Colors
@'e
at'ration Intensit$
RG3
Image
E l " i l i HS& C
-
7/24/2019 colorppt
20/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Eample "anipulating HS& ComponentsEample "anipulating HS& Components
@'e
at'ration Intensit$
RG3
Image @'e at'ration
Intensit$ RG3
Image
C l & P i
-
7/24/2019 colorppt
21/53
Color &mage ProcessingColor &mage Processing
:here are 2 t$&es of color image &rocesses
1. Pse'docolor image &rocess* ;ssigning colors to gra$
%al'es )ased on a s&ecific criterion. Gra$ scale images to )e &rocessed
ma$ )e a single image or m'lti&le images s'ch as m'ltis&ectral images
2. -'ll color image &rocess* :he &rocess to mani&'late real
color images s'ch as color &hotogra&hs.
P d l & P iP d l & P i
-
7/24/2019 colorppt
22/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Pseudocolor &mage ProcessingPseudocolor &mage Processing
Wh$ "e need to assign colors to gra$ scale image
;ns"er* @'man can disting'ish different colors )etter than different
shades of gra$.
Pse'do color false color * In some case there is no Fcolor conce&t
for a gra$ scale image )'t "e can assign Ffalse colors to an image.
& t it Sli i * it Sli i& t it Sli i * it Sli i
-
7/24/2019 colorppt
23/53
&ntensity Slicing or *ensity Slicing&ntensity Slicing or *ensity Slicing
>
=
TyxfC
TyxfCyxg
,(if
,(if,(
2
1
-orm'la*
C1 Color !o. 1
C2
Color !o. 2
:
Intensit$
Co
lor
C1
C2
:+ 41
; gra$ scale image %ie"ed as a 0D s'rface.
& t it Sli i E l& t it Sli i E l
-
7/24/2019 colorppt
24/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
&ntensity Slicing Eample&ntensity Slicing Eample
;n B4ra$ image of a "eld "ith crac>s
;fter assigning a $ello" color to &i?els "ith
%al'e 255 and a )l'e color to all other &i?els.
" lti L l & t it Sli i" lti L l & t it Sli i
-
7/24/2019 colorppt
25/53
"ulti Level &ntensity Slicing"ulti Level &ntensity Slicing
kkk lyxflCyxg
-
7/24/2019 colorppt
26/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
"ulti Level &ntensity Slicing Eample"ulti Level &ntensity Slicing Eample
kkk lyxflCyxg er
:h$roid Phantom.
;fter densit$ slicing into 8 colors
L l t C l + f ti L l t C l + f ti
-
7/24/2019 colorppt
27/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
ray Level to Color +ransformationray Level to Color +ransformation
;ssigning colors to gra$ le%els )ased on s&ecific ma&&ing f'nctions
Red com&onent
Green com&onent
3l'e com&onent
Gra$ scale image
L l t C l + f ti E lray Level to Color +ransformation Eample
-
7/24/2019 colorppt
28/53
(Images from Rafael C.
Gonzalez and Richard
E. Wood, Digital ImageProcessing, 2ndEdition.
ray Level to Color +ransformation Eampleray Level to Color +ransformation Eample
;n B4ra$ image of a
garment )ag "ith a
sim'lated e?&losi%e
de%ice
;n B4ra$ image
of a garment )ag
Color
coded
images
:ransformations
L l t C l + f ti E lray Level to Color +ransformation Eample
-
7/24/2019 colorppt
29/53
(Images from Rafael C.
Gonzalez and Richard
E. Wood, Digital ImageProcessing, 2ndEdition.
ray Level to Color +ransformation Eampleray Level to Color +ransformation Eample
;n B4ra$ image of a
garment )ag "ith a
sim'lated e?&losi%e
de%ice
;n B4ra$ image
of a garment )ag
Color
codedimages
:ransformations
Pse docolor CodingPseudocolor Coding
-
7/24/2019 colorppt
30/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Pseudocolor CodingPseudocolor Coding
Hsed in the case "here there are man$ monochrome images s'ch as m'ltis&ectral
satellite images.
!asics of ,ull Color &mage Processing!asics of ,ull Color &mage Processing
-
7/24/2019 colorppt
31/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
!asics of ,ull-Color &mage Processing!asics of ,ull-Color &mage Processing
2
-
7/24/2019 colorppt
32/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
EampleEample ,ull-Color &mage and ariouis Color Space Components,ull-Color &mage and ariouis Color Space Components
Color image
C
-
7/24/2019 colorppt
33/53
Color +ransformationColor +ransformation
-orm'lation*
[ ].,(.,( yxfTyxg =
f(x,y in&'t color image, g(x,y o't&'t color image
T o&eration on fo%er a s&atial neigh)orhood of (x,y
When onl$ data at one &i?el is 'sedin the transformation, "e
can e?&ress the transformation as*
.,,,( 21 nii rrrTs
= i 1, 2, , n
Where ri color com&onent of f(x,y
si color com&onent of g(x,y
Hse to transform colors to colors.
-or RG3 images, n 0
Eample Color +ransformationEample Color +ransformation
-
7/24/2019 colorppt
34/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Eample Color +ransformationEample Color +ransformation
.,(.,(
.,(.,(
.,(.,(
yxkryxs
yxkryxs
yxkryxs
BB
GG
RR
=
=
=
-orm'la for RG3*
.,(.,( yxkryxs II =
-orm'la for CJ1.
k +.
I H,S
Color ComplementsColor Complements
-
7/24/2019 colorppt
35/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color ComplementsColor Complements
Color com&lement re&laces each color "ith its o&&osite color in the
color circle of the @'e com&onent. :his o&eration is analogo's to
image negati%e in a gra$ scale image.
Color circle
Color &mage SmoothingColor &mage Smoothing
-
7/24/2019 colorppt
36/53
Color &mage SmoothingColor &mage Smoothing
/ "ethods1. Per4color4&lane method*for RG3, C
-
7/24/2019 colorppt
37/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color &mage Smoothing Eample (cont.)Color &mage Smoothing Eample (cont.)
Color image Red
Green 3l'e
Color &mage Smoothing Eample (cont )Color &mage Smoothing Eample (cont )
-
7/24/2019 colorppt
38/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color &mage Smoothing Eample (cont.)Color &mage Smoothing Eample (cont.)
@'e at'ration Intensit$
Color image
@I Com&onents
Color &mage Smoothing Eample (cont )Color &mage Smoothing Eample (cont )
-
7/24/2019 colorppt
39/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color &mage Smoothing Eample (cont.)Color &mage Smoothing Eample (cont.)
mooth all RG3 com&onents mooth onl$ I com&onent of @I
(faster
Color &mage Smoothing Eample (cont )Color &mage Smoothing Eample (cont )
-
7/24/2019 colorppt
40/53
Color &mage Smoothing Eample (cont.)Color &mage Smoothing Eample (cont.)
Difference )et"een
smoothed res'lts from 2methods in the &re%io's
slide.
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color &mage SharpeningColor &mage Sharpening
-
7/24/2019 colorppt
41/53
Color &mage SharpeningColor &mage Sharpening
We can do in the same manner as color image smoothing*
1. Per4color4&lane method for RG3,C
-
7/24/2019 colorppt
42/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color &mage Sharpening Eample (cont.)Color &mage Sharpening Eample (cont.)
Difference )et"een
shar&ened res'lts from 2methods in the &re%io's
slide.
Color SegmentationColor Segmentation
-
7/24/2019 colorppt
43/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color SegmentationColor Segmentation
/ "ethods
1. egmented in @I color s&ace*
; thresholding f'nction )ased on color information in @ and Com&onents. We rarel$ 'se I com&onent for color image
segmentation.
2. egmentation in RG3 %ector s&ace*; thresholding f'nction )ased on distance in a color %ector s&ace.
-
7/24/2019 colorppt
44/53
egmentation in @I model*
K Color is con%enientl$ re&resented in the h'e
image.
K at'ration is 'sed as a mas>ing image inorder to isolate f'rther regions of interest in
the h'e image.
K Intensit$ image is 'sed less fre/'entl$ forsegmentation of color images. 3eca'se it
carries no color information.
Color Segmentation in HS& Color SpaceColor Segmentation in HS& Color Space
-
7/24/2019 colorppt
45/53
Color Segmentation in HS& Color SpaceColor Segmentation in HS& Color Space
@'e
at'ration Intensit$
Color image
1 2
0 #
(Images from Rafael C.
Gonzalez and Richard E.Wood, Digital Image
Processing, 2ndEdition.
Color Segmentation in HS& Color Space (cont.)Color Segmentation in HS& Color Space (cont.)
-
7/24/2019 colorppt
46/53
Color Segmentation in HS& Color Space (cont.)Color Segmentation in HS& Color Space (cont.)
Prod'ct of and
5 6
8
523inar$ thresholding of com&onent
"ith : 1+
@istogram of 6 egmentation of red color &i?els
Red &i?els
(Images from Rafael C.
Gonzalez and Richard E.
Wood, Digital Image
Processing, 2ndEdition.
Color Segmentation in HS& Color Space (cont.)Color Segmentation in HS& Color Space (cont.)
-
7/24/2019 colorppt
47/53
Color Segmentation in HS& Color Space (cont.)Color Segmentation in HS& Color Space (cont.)
Color image egmented res'lts of red &i?els
(Images from Rafael C.
Gonzalez and Richard E.
Wood, Digital Image
Processing, 2ndEdition.
Color Segmentation in R! ector SpaceColor Segmentation in R! ector Space
-
7/24/2019 colorppt
48/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Color Segmentation in R! ector SpaceColor Segmentation in R! ector Space
1. Each &oint "ith (R,G,3 coordinate in the %ector s&ace re&resents
one color.
2. egmentation is )ased on distance thresholding in a %ector s&ace
>=TyxD
TyxDyxg
T
T
,,((if+
,,((if1,(
cc
cc
cT color to )e segmented.
c(x,y RG3 %ector at &i?el (?,$.D(u,v distance f'nction
Eample Segmentation in R! ector SpaceEample Segmentation in R! ector Space
-
7/24/2019 colorppt
49/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
Eample Segmentation in R! ector SpaceEample Segmentation in R! ector Space
Color image
Res'lts of segmentation in
RG3 %ector s&ace "ith :hreshold%al'e
Reference color cTto )e segmented)o?thein&i?elofcolora%erage=
Tc
T 1.25 times the D of R,G,3 %al'es
In the )o?
radient of a Color &mageradient of a Color &mage
-
7/24/2019 colorppt
50/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
radient of a Color &mageg
ince gradient is define onl$ for a scalar image, there is no conce&t
of gradient for a color image. We can9t com&'te gradient of each
color com&onent and com)ine the res'lts to get the gradient of a colorimage.
Red Green 3l'e
Edges
We see
# o)Lects.
We see
2 o)Lects.
radient of a Color &mage (cont.)radient of a Color &mage (cont.)
-
7/24/2019 colorppt
51/53
radient of a Color &mage (cont.)g ( )
Mne "a$ to com&'te the ma?im'm rate of change of a color image
"hich is close to the meaning of gradient is to 'se the follo"ing
form'la* Gradient com&'ted in RG3 color s&ace*
[ ]2
1
2sin22cos((2
1(
+++= xyyyxxyyxx gggggF
( )
=
yyxx
xy
gg
g2tan
2
1 1
222
x
B
x
G
x
R
gxx
+
+
=
222
y
B
y
G
y
R
gyy
+
+
=
y
B
x
B
y
G
x
G
y
R
x
Rgxy
+
+
=
radient of a Color &mage Eampleradient of a Color &mage Eample
-
7/24/2019 colorppt
52/53
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2ndEdition.
M)tained 'sing
the form'lain the &re%io's
slide
'm of
gradients of
each colorcom&onent
Mriginal
image
Difference
)et"een2 and 0
2
0
2 0
g pg p
radient of a Color &mage Eampleradient of a Color &mage Eample
-
7/24/2019 colorppt
53/53
Gradients of each color com&onent
Red Green 3l'e
g pg p