6.837-20 Color 1: Color Vision and Color Spaces

Frédo Durand and Barb Cutler MIT-EECS ColorMany slides courtesy of Victor Ostromoukhovand Leonard McMillanColor Vision 2 Why is the sky blue part II•What do you mean exactly by blue?Color Vision 3 Admin•Quiz on Thursday–1 sheet of notes allowed•Review session tonight 7:30Color Vision 4 Review of last weekColor Vision 5 Monte Carlo Recap•Random rays to sample rendering equation•No meshing required, no special storage•No limitation–On reflectance–On geometry•Can be noisy (or slow )•Advanced–Irradiance cache–Photon mapn1Color Vision 6 You believe you know it all•Color is about spectrum and wavelength•We can get everything from red, green and blue•Well, life is more confusing than that!Color Vision 7 Puzzles about color•How comes a continuous spectrum ends up as a 3D color space•Why is violet “close” to red•Primaries: 3 or 4? Which ones–Red, blue, yellow, green–Cyan and magenta are not “spontaneous” primaries•Color mixing•What is the color of Henry IV’s white horse?Color Vision 14 Plan•Color Vision•Color spaces•Producing color•Color effectsColor Vision 15 Cone spectral sensitivity•Short, Medium and Long wavelengthwavelength0.751.000.500.250.00400500600700SMLColor Vision 16 Cones do not “see” colorswavelength0.751.000.500.250.00400500600700MColor Vision 17 Cones do not “see” colors•Different wavelength, different intensity•Same responsewavelength0.751.000.500.250.00400500600700MColor Vision 18 Response comparison•Different wavelength, different intensity •But different response for different coneswavelength0.751.000.500.250.00400500600700SMLColor Vision 19 von Helmholtz1859: Trichromatictheory•Colors as relative responses(ratios)VioletBlueGreenYellowOrangeRedShort wavelength receptorsMedium wavelength receptorsLong wavelength receptorsReceptor ResponsesWavelengths (nm)400 500600700VioletBlueGreenYellowOrangeRedColor Vision 20 Cones distribution•In the retina•LMS 40:20:1•No S (blue) in retina centerColor Vision 21 Metamers•Different spectrum•Same response Wavelength (nm) Wavelength (nm) 400 480 500 580 600 700 400 500 600 700 INTENSITY INTENSITYColor Vision 22 Color matching•Reproduce the color of a test lampwith the addition of 3 primary lightsColor Vision 23 Metamerism & light source•Metamersunder a given light source•May not be metamersunder a different lampColor Vision 24 Color blindness•Dalton •8% male, 0.6% female•Genetic•Dichromate (2% male)–One type of cone missing–L (protanope), M (deuteranope), S (tritanope)•Anomalous trichromat–Shifted sensitivityColor Vision 26 Color blindness test•Maze in subtle intensity contrast•Visible only to color blinds•Color contrast overrides intensity otherwiseColor Vision 27 Plan•Color Vision–Cone response, trichromats–Opponent theory–Higher-level•Color spaces•Producing color•Color effectsColor Vision 28 Remember von Helmholtz•Colors as relative responses(ratios)VioletBlueGreenYellowOrangeRedShort wavelength receptorsMedium wavelength receptorsLong wavelength receptorsReceptor ResponsesWavelengths (nm)400 500600700VioletBlueGreenYellowOrangeRedColor Vision 29 Hering1874: Opponent Colors+0-+0-+0-Red/GreenReceptorsBlue/YellowReceptorsBlack/WhiteReceptors•Hypothesis of 3 types of receptors: Red/Green, Blue/Yellow, Black/White•Explains well several visual phenomena•Hypothesis of 3 types of receptors: Red/Green, Blue/Yellow, Black/White•Explains well several visual phenomenaColor Vision 30 Dual Process Theory•The input is LMS•The output has a different parameterization:–Light-dark–Blue-yellow–Red-greenSLMGYWhBkBRTrichromaticStageOpponent-ProcessStageColor Vision 31 Color opponents wiring•Sums for brightness•Differences for color opponentsB+Y-W+B-R +G-MLSML++++++--Y+B-B+W-G+R-MLSML++++----S-M-LS+M+LL-M-S+M+L-S-M-LM-LColor Vision 32 Simultaneous contrast•In color opponent direction•Center-surround R+GGGRBBYYYBYYBRRGGGRBBYColor Vision 33 Land RetinexColor Vision 34 Simultaneous Color ContrastColor Vision 35 After-ImageColor Vision 36 Opponent ColorsImageImageAfterimageAfterimageColor Vision 37 Opponents and image compression•JPG, MPG•Color opponents instead of RGB•Compress color more than luminanceColor Vision 38 Plan•Color Vision–Cone response, trichromats–Opponent theory–Higher-level•Color spaces•Producing color•Color effectsColor Vision 39 Color reparameterization•The input is LMS•The output has a different parameterization:–Light-dark–Blue-yellow–Red-green•A later stage may reparameterize:–Brightness or Luminance or Value–Hue–SaturationSLMGYWhBkBRSL (or B)HColor Vision 40 Hue Saturation Value Courtesy of Stephen Palmer, VISION SCIENCE, and the MIT Press.Used with permission.Color Vision 41 Hue Saturation Value•One interpretation in spectrum space•Not the only onebecause of metamerism•Dominant wavelength (hue)•Intensity•Purity (saturation)Color Vision 42 Color categories•Prototypes•Harder to classify colors at boundaries Hue 0 Degree of Membership 1 Focal Blue Focal Green (Red) Blue Green Yellow Red (Blue) Focal Yellow Focal RedColor Vision 43 Plan•Color Vision•Color spaces•Producing color•Color effectsColor Vision 44 Color spaces•Human color perception is 3 dimensional•How should we parameterize this 3D space•Various constraints/goals–Linear parameterization–Close to color technology–Close to human perception–StandardColor Vision 45 The root of all evil•Cone responses are not orthogonal (they overlap)•To change the M response without changing the L one, we need negative lightwavelength0.751.000.500.250.00400500600700SMLOrthogonal basis (color matching function)Color Vision 46 Color Matching Problem•Some colors cannot be produced using only positively weighted primaries•E.g. primaries: pure wavelength–650, 530, 460•Some colors need negative amounts of primaries•Analysis spectrum hasnegative lobesColor Vision 47 Color Matching Problem•Some colors cannot beproduced using only positively weighted primaries•Solution: add light on the other side!Color Vision 48 Color Matching Problem•Some colors cannot be produced using only positively weighted primaries•Some tradeoff must be found between negative lobes in analysis vs. synthesis•In 1931, the CIE (Commission Internationalede L’Eclairage) defined three new primaries•Called X, Y , Z,–with positive color matching functionsColor Vision 49 CIE color space•Can think of X, Y , Z as coordinates•Linear transform from RGB or LMS x R z y 3.24 G -0.97 B 0.06 -1.54 1.88 -0.20 -0.50 0.04 -1.06 ( ( ( ( XYZ ( ( = X 0.41 Y 0.21 Z 0.02 0.36 0.72 0.12 0.18 0.07 0.95 ( ( ( ( RGB ( ( =Color Vision 50 CIE color space•Odd-shaped cone contains visible colors–Note that many points in XYZ do not correspond to visiblecolors! x R z y 3.24 G -0.97 B 0.06 -1.54 1.88 -0.20 -0.50 0.04 -1.06 ( ( ( ( XYZ ( ( = X 0.41 Y 0.21 Z 0.02 0.36 0.72 0.12 0.18 0.07 0.95 ( ( ( ( RGB ( ( =Color Vision 51 CIE color space•Objective, quantitative color descriptions–Dominant wavelength:•Wavelength “seen” (corresponds to Hue)–Excitation purity:•Saturation, expressed objectively–Luminance:•Intensity•Chromaticity (independent of luminance):–normalize against X + Y + Z:Color Vision 52 CIE color space•Spectrally pure colors lie along boundary•Note that some huesdo not correspond to a pure spectrum (purple-violet)•Standard white light (approximates sunlight) at CC 0.1 0.1 0.2 400 480 490 500 510 Green 520 540 560 580 Yellow Red Purple Blue Cyan 600 700 0.2 0.3 0.3 0.4 0.4 0.5 0.5 0.6 0.6 0.7 0.8 x y 0.7 0.8 0.9 C Hunt, R. W. G. The Reproduction of Colour. John Wiley & Sons Incorporated. September 2004. ISBN: 0-470-02425-9. Image adapted from:Color Vision 53 CIE color space•Match color at some point A•A is mix of white C, spectral B!•What is dominant wavelength of A?•What is excitation purity (%) of A?–Move along AC/BCC 0.1 0.1 0.2 400 480 490 500 510 520 540 560 B A D E C F G 580 600 700 0.2 0.3 0.3 0.4 0.4 0.5 0.5 0.6 0.6 0.7 x y 0.7 0.8 Hunt, R. W. G. The Reproduction of Colour. John Wiley & Sons Incorporated. September 2004. ISBN: 0-470-02425-9. Image adapted from:Color Vision 54 XYZ vs. RGB•Linear transform•XYZ is more standardized•XYZ can reproduce all colors with positive values•XYZ is not realizable physically !!–What happens if you go “off” the diagram–In fact, the orthogonal (synthesis) basis of XYZ requires negative values.Color Vision 55 Perceptually Uniform Space: MacAdam•In color space CIE-XYZ, the perceived distance between colors is not equal everywhere•In perceptually uniform color space, Euclidean distances reflect perceived differences between colors•MacAdamellipses (areas of unperceivable differences) become circles 560 550 540 530 520 510 490 480 470 460 450 .05 x y .10 .15 .20 .45 .50 .55 .60 .65.25 .25 .20.15 .10 .05 .35 .40 .45 Spectrum locus.50 .55 .60 .65 .70 .30 400 nm 700 nm 650 620 610 600 590 580 570 Purple line Wyszecki, G. and W. S. Stiles. Color science: Concepts and Methods, Quantitative Data and Formulae. Wiley-Interscience, 2nd ed. July 2000. ISBN: 0471399183. Image adapted from:Color Vision 56 CIE-LABSource: [Wyszeckiand Stiles ’82] Image adapted from Wyszecki, G., and W. S. Stiles,Color science: Concepts and methods, quantitative data and formulae. Wiley-Interscience; 2nd Edition. July 2000. ISBN: 0471399183Color Vision 57 Perceptually Uniform Space MunsellMunsellColor SpaceHueValueChromawww.munsell.communsell.comColor Vision 58 Color response linear subspace•Project the infinite-D spectrum onto a subspace defined by 3 basis functions•We can use 3x3 matrices to change the colorspace–E.g. LMS to RGB–E.g. RGB to CIE XYZColor Vision 59 Color response and RGB or LMS•Project the infinite-D spectrum onto a subspace defined by 3 basis functions•Small problem: this basis is NOT orthogonal•What does orthogonal mean in our case?•Second problem: the orthogonal basis is NOT physically realizableColor Vision 60 Color response and RGB or LMS•Project the infinite-D spectrum onto a subspace defined by 3 basis functions•Small problem: this basis is NOT orthogonal•What does orthogonal mean in our case?•Second problem: the orthogonal basis is NOT physically realizableColor Vision 61 Munsellbook of colors•Perceptually uniformwww.munsell.com