Rochester Institute of Technology RIT Scholar Works Presentations and other scholarship 1-18-25 Color appearance in image displays Mark Fairchild Follow this and additional works at: http://scholarworks.rit.edu/other Recommended Citation Fairchild, Mark, "Color appearance in image displays" (25). Accessed from http://scholarworks.rit.edu/other/13 This Presentation is brought to you for free and open access by RIT Scholar Works. It has been accepted for inclusion in Presentations and other scholarship by an authorized administrator of RIT Scholar Works. For more information, please contact ritscholarworks@rit.edu.
Color Appearance in Image Displays Mark D. Fairchild RIT Munsell Color Science Laboratory ISCC/CIE Expert Symposium 75 Years of the CIE Standard Colorimetric Observer Ottawa 26 O Canada
Image Colorimetry Device Dependent (e.g. RGB, CMYK) Device Independent (e.g. XYZ, L*a*b*) Viewing-Conditions Independent (e.g. JCh) Color Appearance Viewing-Conditions Independent But Spatially Localized
Image Appearance Spatial, Temporal & Image Quality Questions Remain... Which degraded image is better? And by how much? XYZ to CIELAB Tristimulus Values Amounts of Three Primaries Required to Match a Color Specifies the Stimulus of Color CIELAB Aims to Begin Describing Differences & Appearance
CIELAB to CIECAM97s CIELAB Adaptation & Response Compression Input: Stimulus & White Point Output: ~ Appearance Correlates Lightness, Chroma, Hue CIECAM97s Aims to be More Accurate & Comprehensive than CIELAB CIECAM: 97s to 2 More Appearance Phenomena Background, Surround, etc. Stimulus, White, Luminance, Other Parts of Viewing Field More Accurate Appearance Correlates Brightness, Lightness, Colorfulness, Chroma, Saturation, Hue CIECAM2: An Evolutionary Enhancement
CIECAM2 Accurate Adaptation & Appearance Scales Simple Stimulus, Background, Surround Images: When Reproduced at Same Scale CIECAM2 Successes Alive After 4+ Years Accurate CAT2 Display Brightness/Colorfulness Perceived Gamut Volumes Practical Color Management (e.g. Microsoft) Color Differences
Beyond CIECAM2 Color Appearance Spatial Vision Temporal Vision Image Appearance Modeling icam: An Example
Past Results Image Quality: Sharpness & Contrast HDR Rendering: Local Adaptation HDR Video: Adaptation Time Course Image Quality 1.2 (a) 14 1 12 Im Model Prediction.8.6.4 Im Model Prediction 1 8 6 4.2 2-5 -4-3 -2-1 1 2 Perceived Difference -4. -3. -2. -1.. 1. 2. Perceived Contrast Image Difference Prediction (Sharpness Data) Image Difference Prediction (Contrast Data)
HDR Rendering Video Rendering
Recent Research HDR Rendering Psychophysics Perceived Color Gamut Volumes Surround Effects Noise Adaptation Orthogonal Opponent-Colors Dimensions HDR Psychophysics Fig. 3 Average preference scores for 12 scenes (color images) (The algorithms are labeled as Retinexbased filters (R), Sigmoid function (S), Histogram adjustment (H), icam (I), Photographic reproduction (P), and Bilateral filter (B). The same labels are used in this article).
HDR Accuracy Fig. 2 Experimental scenes: (a) window (b) breakfast (c) desk Fig. 14 Overall accuracy scores for HDR rendering algorithms Perceived Gamuts
Image Examples Image Examples
Surround Figure 1, Demos for surround lab. Image is shown in different surround condition. Relative Lum Normalized Lum 1.5 1.5.5 1 1.5 2 1.5 1 1.5 1.5 1.5.5 1 1.5 2 1.5 1 1.5 1.5 1.5.5 1 1.5 2 1.5 1 1.5 1.5 1.5.5 1 1.5 2 1.5 1 1.5 1.5 1.5.5 1 1.5 2 1.5 1 1.5 Figure 5, Relationship between the average surround luminance and image contrast. Each column shows the plot for each scene. The first row shows the scene; the second row shows the average surround luminance (relative to maximum LCD luminance) vs. gamma for each scene; in the last row, the relative surround luminance is normalized to the average luminance of three different gammas ( = 1/1.3, = 1, and = 1.3 ) for each scene. Noise Adaptation
Noise Adaptation Noise Adaptation
.9 Noise Adaptation MF Random Adapt MF Horizontal Adapt MF Vertical Adapt GJ Random Adapt GJ Horizontal Adapt GJ Vertical Adapt.8.3 Random Adapt Horizontal Adapt Vertical Adapt Visible Contrast.7.6 Model-Equal Random Contrast.25.5.2..1.2.3.4 Adapting Contrast.4..1.2.3.4 Adapting Contrast Orthogonal Opponency C C $%&'()"*+,-./1+2),$(345)6773(%&87978)78%(:96%(79+;:)23;::%4);%3;, A $%&'()"*<,.=>?@A1+2),$(345)6773(%&879?9@9+;:A, A C B C B PCA3 PCA3 $%&'()"*2,=-B1+2),$(345)6773(%&879=9-9+;:B PCA2 PCA2 PCA1 A PCA1 $%&'()"*:,@.<.(1+2),$3(45)6773(%&879@9.<9+;:.(, A B B For Y average =.29 # # # " V 1 V 2 V 3 $ '.48.196.9998 $ X $ & # & # & & = #'.9983.578 '.59& * # Y & & # % "'.579 '.9981.193& # % " Z & %
Ongoing Projects Image Size Improved HDR Rendering HDR Photographic/Appearance Survey Spectral Adaptation Modeling Transformability of Primaries CIECAM2 & IPT E Gamuts & Brilliance Observer Metamerism & Full Standard Observer Surround Color Color Curiosity Shop Image Size 43.6 (mm) 87.2 (mm) 174.3 (mm) 261.5 (mm)
Improved HDR HDR Survey
Gamuts & Brilliance 3 cd/m 2 :.3 cd/m 2 (1,:1 Contrast) Spectral Adaptation Illuminant/Source WL (nm) Illuminant/Source WN (cm -1 ) Reflectance WL (nm) Reflectance WN (cm -1 ) 14. Define Blur WN (cm -1 ) Stimulus WN (cm -1 ) Median CIELAB Color Difference 12. 1. 8. 6. 4. 2. A D75 TL84 Hor CWF Equal-Energy Illuminant WN (cm -1 ) (1-D) Blurred Illuminant/Source WN (cm -1 ) + (D) Adapting Stimulus WN (cm -1 ). Spectral CAT2 CIELAB Constancy Adaptation Model Adapted Stimulus WN (cm -1 ) Adapted Ill. E Reflectance WL (nm) Colorimetry Illuminant E XYZ CIELAB
CIECAM2 E E94-Type Weighted Color Difference Equation In CIECAM2 JCh and IPT(cylindrical) Performance Comparable to DE2 Useful Default Viewing Conditions More Testing and Description to Come Transformability TC1-56, Improved Colour Matching Functions, M. Brill Match 2 Whites, Each with Two Sets of RGB Primaries Many Replicates for a Few Observers A Collection of Observers Statistically Meaningful Test of Transformability of Primaries One of Several Labs Thanks to Irena Fryc @ NIST
2 ) (L*) b* -a* a* -b* a*.25.2.15.1.5 b* L* b* L* 1.. -1. 1.. -1. Power CRT Gray Print. 38 48 58 68 78 Wavelength (nm) HARD-COPY a*. 1. a*. 1. a* Metameric Matching Color Reproduction Media 2 Observers 7 Colors (CMYKRGB) 2 Media (Print, Transparency) L* Nimeroff et al. CIE Pub. 8 Intra-observer b* Inter-observer Sample: Cyan Transparency.5 1. 1.5 L-Cone Peak Density.75 1. 1.25 Lens Peak Density.33 1. 1.67 M-Cone Peak Density Funding: NSF-NYS/IUCRC & NYSSTF CAT CEIS Visual Experiment: Rick Alfvin and Jason Gibson.2 1. 1.8 Macula Peak Density.33 1. 1.67 Lens Peak Density Experimental Results: R.L. Alfvin and M.D. Fairchild, Observer Variability in Metameric Color Matches using Color Reproduction Media, Color Res. Appl. 22, in press (1997). CMF Model: A.D. North and M.D. Fairchild, Measuring Color Matching Functions Part I, Color Res. Appl. 18, 155-162 (1993). Visual Data Starting Point: V.C. Smith and J. Pokorny, Chromatic Discrimination Axes, CRT Phosphor Spectra, and Individual Variation in Color Vision, J. Opt. Soc. Am. 12, 27-35 (1995). Observed Predicted Obs. Metamerism... Modeling Observer Metamerism through Monte Carlo Simulation Abstract: Monte Carlo Experiment: 1, Sets of Color Matching Functions Generated 1996 OSA Poster Improved Model More Observers Full Expression of Nimeroff Mean- Covariance System Metameric color matches depend on the observer s color matching functions. Data were collected on observer variability in typical metameric matches. A Monte Carlo simulation, using a model of color matching functions and physiological data, was performed to derive a complete colorimetric system capable of predicting inter-observer variability in addition to mean color matches. Visual Experiment: CRT Typical Results: Radiance (w/sr*m Inter-Observer Variability Observed and predicted (previously published models) covariance ellipses. Predictions are inadequate. Monte Carlo Model: x' (") = 1 #k 1x $ lens (") 1 #k 2 x $ macula (") [ k 3 x L(") + k 4 x M(") + k 5 x S(") ] y' (") = 1 #k 1y$ (") lens 1 #k 2 y$ macula(") [ k 3 yl(") + k 4yM(") + k 5 ys(") ] z' (") = 1 #k 1z$ (") lens 1 #k 2 z$ macula (") [ k 3 z L(") + k 4 z M(") + k 5 z S(") ] L-Cones: 6% Smith & Pokorny 4% Shifted -4nm (In Wave#) k Coefficeints Fitted to CIE 1931 Standard Colorimetric Observer M-Cones: 88% Smith & Pokorny 12% Shifted +4nm (In Wave#) Acknowledgements / References: S-Cones: 1% Smith & Pokorny Mean and Covariance Functions Established Standard Error Propoagation to CIELAB Covariance Matrices for Observed Metamers Predicted Covariance Dependent upon Metemeric Properties Monte Carlo Results: Gray Print: 1, Color Matching Functions Blue Transparency: 1, Color Matching Functions Gray Print: 3 Sets of 2 Color Matching Functions Conclusions: Observer Variability in Practical Color Matching is Significant Previously Published Techniques Underpredict Variability A Monte Carlo Model Produced Better Results Further Data and Model Refinement are Required Surround Color Before Model After
Curiosity Conclusions Image Appearance Modeling is a Natural Extension of Color Appearance Modeling Enabled by Recent Technology There are Many Questions of Fundamental and Applied Color Science that Build Together to Address Image Appearance Plenty of Exciting Challenges Lie Ahead
Thank You... The work discussed in this paper has been funded by a variety of corporate, government, and institutional sponsors. See <mcsl.rit.edu/about/sponsors.php> for a full listing.