Color Image Processing................ppt

this is our Computer Graphics note

1. Computer Graphics
Chapter Seven
Color Image Processing

2. TheuseofcolorImageProcessingismotivated
bytwoprincipalfactors:
Colorisapowerfuldescriptor
Humanscandistinguishbetweenthousandsofcolor
shadesandintensitiescomparedtoaboutonlytwo
dozenshadesofgray
Color Image Processing

3. Full Color Processing vs Pseudo-Color Processing
InFullColorProcessingtheimageinquestiontypically
areacquiredwithaFull-Colorsensore.g.ColorTV
cameraorColorScanner
InPseudo-colorProcessingtheproblemisofassigninga
colortoaparticularmonochromeintensityorarangeof
intensities

4. Color Spectrum

5. Electromagnetic Spectrum

6. Physical Background
•Visiblelight:anarrowbandof
electromagneticradiation→
380nm(blue) -780nm
(red)
•Wavelength:Eachphysically
distinctcolourcorrespondsto
atleastonewavelengthinthis
band.

7. Color Fundamentals
Thecolorsthathumans andsome
animalsperceiveinanobjectare
determinedbythenatureoflight
reflectedfromtheobject

8. Achromatic vs Chromatic Light
Achromatic(voidofcolor)Light:Itsonly
contributeisits‘Intensity’oramount
ChromaticLight:spanstheelectromagnetic
spectrumfromapproximately400to700nm

9. Quantities for description of quantity of Chromatic
Source of Light
Threebasicquantitiesareusedtodescribethequantityofa
chromaticsourceoflight:
Radiance
Luminance
Brightness

10. Radiance
ThetotalamountofEnergythatflowsfromaLight
Source
ItismeasuredinWatts

11. Luminance
Luminancegivesameasureofamountofenergyan
observerperceivesfromalightsource(measuredin
Lumens(lm))
Forexamplelightemittedfromasourceoperatingin
InfraredregionofSpectrumcouldhavesignificant
energy(Radiance)butahumanobserverwillhardly
perceiveitsoluminanceiszero.

12. Brightness
Itisasubjectivemeasure
Itembodiestheachromaticnotionofintensityandisoneof
thekeyfactorsindescribingcolorsensation

13. Human Perception
Detailedexperimentalevidenceshasestablishedthatthe6
to7millionconesinthehumaneyecanbedividedinto
threeprincipalsensingcategories,correspondingroughlyto
red,greenandblue
Approximately65%ofallconesaresensitivetoRedLight,
33%aresensitivetoGreenLightandabout2%are
sensitivetoBlueLight(mostsensitive)

14. Human Perception
DuetotheseabsorptioncharacteristicofHumanEye
colorsareseenasvariablecombinationsoftheso-called
‘PrimaryColors’Red,GreenandBlue
Theprimarycolorscanbeaddedtoproducesecondary
colorsofLight
Magenta(Red+Blue)
Cyan(Green+Blue)
Yellow(Red+Green)

15. Absorption of Light by red, green and blue
cones in Human Eye
Mixingthethreeprimariesorasecondarywithitsopposite
primarycolorsintherightintensitiesproduceswhitelight

16. Primary Color of Light vs Primary Color of Pigments
Red,GreenandBlueColorsarePrimaryColorsofLight
InPrimaryColorofPigmentsaprimarycolorisdefinedas
theonethatsubtractsorabsorbsaprimarycolorofLight
andreflectsortransmitstheothertwo
ThereforethePrimaryColorsofPigmentsareMagenta,
CyanandYellowandsecondarycolorsareRed,Greenand
Blue
Apropercombinationofthreepigmentprimariesora
secondarywithitsoppositeprimaryproducesBlack
ColorTelevisionReceptionisanexampleoftheadditive
natureofLightColors

18. Tri-Stimulus Values
TheamountofRed,GreenandBlueneededtoforma
particularcolor(denotedbyX,YandZ)
Acoloristhenspecifiedbyits“Tri-chromaticCoefficients”
•Thus x+y+z=1

19. Chromaticity Diagram
Anotherapproachforspecifyingcolorsistouse
chromaticitydiagram
Showscolorcompositionsasafunctionofx(red)and
y(green)
Foranyxandythecorrespondingvalueofz(blue)can
beobtainedas
z=1-x-y

20. Chromaticity Diagram

21. Chromaticity Diagram
Todeterminetherangeof
colorsthatcanbeobtained
fromthe3givencolorsinthe
CD, we simply draw
connectinglinestoeachof
thethreecolorpoints.
Theresultisatriangleand
anycolorinsideatriangleis
produced by various
combinationsofthethree
initialcolors.
Thetriangleshowsatypical
rangeofcolors(calledthe
colorgamut)producedby
RGBmonitor

22. Color Models
Thepurposeofacolormodel(alsocalledColorSpaceorColor
System)istofacilitatethespecificationofcolorsinsome
standardway
Acolormodelisaspecificationofacoordinatesystemanda
subspacewithinthatsystemwhereeachcolorisrepresentedby
asinglepoint
ColorModels
RGB(Red,Green,Blue)
CMY(Cyan,Magenta,Yellow)
HSI(Hue,Saturation,Intensity)
YIQ(Luminance,In phase,Quadrature)
YUV(Y'standsforthelumacomponent (thebrightness)andUandV
arethechrominance(color)components)

23. RGB Model
Eachcolorisrepresented
initsprimarycolor
components Red,Green
andBlue
Thismodelisbasedon
Cartesian Coordinate
System

24. RGB Model
Inthismodel,theprimarycolorsarered,green,andblue.Itis
anadditivemodel,inwhichcolorsareproducedbyadding
components,withwhitehavingallcolorspresentandblack
beingtheabsenceofanycolor.
Thisisthemodelusedforactivedisplayssuchastelevision
andcomputerscreens.
TheRGBmodelisusuallyrepresentedbyaunitcubewithone
cornerlocatedattheoriginofathree-dimensionalcolor
coordinatesystem,theaxesbeinglabeledR,G,B,andhaving
arangeofvalues[0,1].Theorigin(0,0,0)isconsidered
blackandthediagonallyoppositecorner(1,1,1)iscalled
white.Thelinejoiningblacktowhiterepresentsagrayscale
andhasequalcomponentsofR,G,B.

25. RGB Color Cube
Thetotalnumberofcolorsina24Bitimageis
(2
8
)
3
=16,777,216(>16million)

26. Generating RGB image

27. CMY and CMYK Color Model
Cyan,magenta,andyellowarethesecondarycolorswithrespecttothe
primarycolorsofred,green,andblue.However,inthissubtractivemodel,
theyaretheprimarycolorsandred,green,andblue,arethesecondaries.
Inthismodel,colorsareformedbysubtraction,whereaddingdifferent
pigmentscausesvariouscolorsnottobereflectedandthusnottobeseen.
Here,whiteistheabsenceofcolors,andblackisthesumofallofthem.
Thisisgenerallythemodelusedforprinting.
Mostdevicesthatdepositcolorpigmentsonpaper(suchasColorPrinters
andCopiers)requiresCMYdatainputorperformRGBtoCMYconversion
internally
C
M
Y
R
G
B
=
1.00
1.00
1.00
-

28. CMY and CMYK Color Model
CMYisaSubtractiveColorModel
EqualamountsofPigmentprimaries(Cyan,Magentaand
Yellow)shouldproduceBlack
Inpracticecombiningthesecolorsforprintingproduces
a“Muddy-Black”color
Soinordertoproduce“True-Black”afourthcolor
“Black”isaddedgivingrisetoCMYKmodel

29. CMY Color Model

30. CMY Color Model

31. HSI Color Model
Hue(dominantcolourseen)
Wavelengthofthepurecolourobservedinthesignal.
Distinguishesred,yellow,green,etc.
Morethe400huescanbeseenbythehumaneye.
Saturation(degreeofdilution)
Inverseofthequantityof“white”presentinthesignal.A
purecolourhas100%saturation,thewhiteandgrey
have0%saturation.
Distinguishesredfrompink,marinebluefromroyalblue,
etc.
About20saturationlevelsarevisibleperhue.
Intensity
Distinguishesthegraylevels.

32. HSI Color Model
SeparatesoutintensityIfromthecoding
Twovalues(Hue&Saturation)encodechromaticity
Intensityencodemonochromepart.
Hueandsaturationofcolorsrespondcloselytotheway
humansperceivecolor,andthusthismodelissuitedfor
interactivemanipulationofcolorimages.

34. Properties of HSI (HSV)
HueHisdefinedbyanangle
SaturationSmodelsthepurityofthecolor
 I=(R+G+B)/3

35. Conversion from RGB to HSI
GivenanimageinRGBcolorformat,theHcomponentofeach
RGBpixelisobtainedusingtheequation:

36. Conversion from HSI to RGB

37. Conversion from HSI to RGB

38. Pseudo-Color (False Color) Image
Processing
Pseudo-colorImageProcessingconsistsofassigningcolors
tograylevelsbasedonspecificcriterion
Generally,theeyecannotdistinguishmorethanabout50
graylevelsinanimage.Thussubtledetailcaneasilybelost
inlookingatgrayscaleimages.Toenhancevariationsin
graylevelandmakethemmoreobvious,grayscaleimages
arefrequentlypseudo-colored,whereeachgrayscale
(generallyatleast256levelsformostdisplays)aremapped
toacolorlevelthroughaLUT.Theeyeisextremely
sensitivetocolorandcandistinguishthousandsofcolor
valuesinapicture.

39. Pseudo-Coloring using LUT
CLUT(Colorlookuptable)::Amappingofapixelvaluetoa
colorvalueshownonadisplaydevice.
•Forexample,inagrayscaleimagewithlevels0,1,2,3,
and4,pseudocoloringisacolorlookuptablethatmaps0
toblack,1tored,2togreen,3toblue,and4towhite.

40. •Thetechniqueofintensityslicingordensityslicingorcolor
codingisoneofthesimplestexampleofPseudo-colorimage
processing
Intensity Slicing

41. •TheGrayScale[0,L-1]isdividedintoLlevels;wherel
0
representsBlack(f(x,y)=0)andl
L-1representswhite
(f(x,y)=L-1)
•SupposethatPplanesperpendiculartotheintensityaxis
aredefinedatlevelsl
1,l
2…..,l
p
•Thenassumingthat0<P<L-1thePplanespartitionthe
grayscaleintoP+1intervals,V
1,V
2…….V
p+1
Intensity Slicing

42. Intensity Slicing
•Grayleveltocolorassignmentsaremade
accordingtotherelation:
f(x,y)=c
kiff(x,y)€v
k
•Wherec
kisthecolorassociatedwiththekth
intensityintervalv
kdefinedbythepartition
planesatl=k-1andl=k

43. An Alternative View of Intensity Slicing

46. Basics of Full Color Image Processing
•Full color image processing fall into 2 categories.
•In1
st
categoryweprocesseachcomponentimage
individuallyandthenformacompositeprocessedcolor
imagefromtheindividuallyprocessedcomponent.
•In2
nd
categoryweworkwithcolorpixelsdirectly.
Becausefullcolorimageshaveatleastthree
components,colorpixelsarereallyvectors.
•LetcrepresentanarbitraryvectorinRGBcolorspace:

47. Basics of Full Color Image Processing
•Colorcomponentsarethefunctionofco-ordinates(x,y)so
wecanwriteitas:
•ForanimageofsizeMxNthereareMNsuchvectors,
c(x,y),forx=0,1,2,…,M-1;y=0,1,2,…,N-1

48. Basics of Full Color Image Processing

49. Color Transformations
•Colortransformationcanberepresentedbythe
expression::
g(x,y)=T[f(x,y)]
f(x,y): inputimage
g(x,y):processed(output)image
T[*]:anoperatoronfdefinedoverneighborhoodof(x,y).
Thepixelvaluesherearetripletsorquartets(i.egroupof3
or4values)

50. Color Transformations
• Si=Ti(r1,r2,…,rn)i=1,2,3,….n
riandSiarevariablesdenotingthecolorcomponentsoff(x,y)and
g(x,y)atanypoint(x,y).
nisthenoofcolorcomponents
{T1,T2,…..,Tn}isasetoftransformationorcolormappingfunctions.
• Note that n transformations combine to produce a single
transformation T

51. Color Transformations
•Thecolorspacechosendeterminethevalueofn.
•IfRGBcolorspaceisselectedthenn=3&r1,r2,r3denotesthered,
blueandgreencomponentsoftheimage.
•IfCMYKcolorspaceisselectedthenn=4&r1,r2,r3,r4denotesthe
cyan,hue,magentaandblackcomponentsoftheimage.
•Supposewewanttomodifytheintensityofthegivenimage
usingg(x,y)=k*f(x,y)where0<k<1

52. Color Transformations
•InHSIcolorspacethiscanbedonewiththesimple
transformation s3=k*r3
wheres1=r1ands2=r2
Onlyintensitycomponentr3ismodified.
•InRGBcolorspace3componentsmustbetransformed:
si=k*ri i=1,2,3.
•Usingk=0.7theintensityofanimageisdecreasedby30%

53. Color Transformations

54. Color Complements
•ThehuesoppositetooneanotherontheColorCirclearecalled
Complements.
•ColorComplementtransformationisequivalenttoimagenegative
inGrayscaleimages

55. Color Complements

56. Color Slicing
•Highlightingaspecificrangeofcolorsinanimageisusefulfor
separatingobjectsfromtheirsurroundings.
•Displaythecolorsofinterestsothattheyaredistinguishedfrom
background.
•Onewaytosliceacolorimageistomapthecoloroutsidesome
rangeofinteresttoanonprominentneutralcolor.

57. Histogram Processing
•Colorimagesarecomposed
ofmultiplecomponents,
howeveritisnotsuitableto
process each plane
independentlyincaseof
histogramequalization.This
resultsinerroneouscolor.
•Amorelogicalapproachis
to spread the color
intensitiesuniformly,leaving
thecolorsthemselves(hue,
saturation)unchanged.
•HSIapproachisideally
suitedtothistypeof
approach.

58. Color Image Smoothing
•Colorimagescanbesmoothedinthesamewayasgrayscale
images,thedifferenceisthatinsteadofscalargraylevelvalueswe
mustdealwithcomponentvectorsofthefollowingform:
•TheaverageoftheRGBcomponentvectorinthisneighborhood
is:

59. Color Image Smoothing
•Werecognizethecomponentsofthisvectorasthescalar
imagesthatwouldbeobtainedbyindependentlysmoothing
eachplaneofthestartingRGBimageusingconventionalgray
scaleneighborhoodprocessing.
•Thusweconcludethatsmoothingbyneighborhoodaveraging
canbecarriedoutonapercolorplanebasis.

60. Color Image Smoothing

61. Color Image Smoothing

62. Color Image Sharpening

63. Noise in Color Images
•Noiseincolorimagescanberemovedthroughvariousnoise
modelswhichweuseinImageRestorationincasethenoise
contentofacolorimagehasthesamecharacteristicsineach
colorchannel.
•Butitispossibleforcolorchannelstobeaffecteddifferentlyby
noisesointhiscasenoiseareremovedfromtheimageby
independentlyprocessingeachplane
•Removenoisebyapplyingsmoothingfilters(e.ggaussian,
average,median)toeachplaneindividuallyandthencombine
theresult.

64. Noise in Color Images

65. Color Image Compression
•Compressionistheprocessof
reducingoreliminatingredundant
and/orirrelevantinformation
•Acompressed imageisnot
directlydisplayableitmustbe
decompressed beforeinputtoa
colormonitor.
•Incaseifinacompressedimage1
bitofdatarepresents230bitsof
dataintheoriginalimage,then
compressed imagecould be
transmittedoverinternetin1
minuteascomparedtooriginal
imagewhichwilltake4hoursto
transmit.

66. Discussion points
1.what are the different color model
and their application in computer
graphics?
2.Describe the common practices that
colors can be used in computer
graphics?
3.Explain noise in image processing
and how it can be eliminated?

67. Any question