Using Color Appearance Models in Device-Independent Color Imaging
The Problem Jackson, McDonald, and Freeman, Computer Generated Color, (1994). MacUser, April (1996)
The Solution Specify Color Independent of Devices* and Viewing Conditions** * Requires Colorimetry ** Requires Color Appearance Model
Outline of Presentation Basic and Advanced Colorimetry Color Reproduction Evolution Color Appearance Phenomena Chromatic Adaptation Models Color Appearance Models (RLAB) Tests of Color Appearance Models
Basic and Advanced Colorimetry According to Wyszecki (1973): Basic Colorimetry involves the specification of color matches (i.e., nominal scaling, CIE XYZ). Advanced Colorimetry involves the specification of color differences and color appearance (i.e., interval and ratio scaling, CIELAB and beyond).
Use of XYZ CIE XYZ Tristimulus Values Represent a Nominal Scale Specify Matches (single viewing condition) Not Color Differences Not Color Appearance
Chromaticity Diagram 530 550 580 700 480 470 430
Where's White? 1.0 0.8 Tungsten Print White y 0.6 D65 Monitor White 0.4 0.2 0.0 0.0 0.2 0.4 x 0.6 0.8 1.0
Color Differences Hunter and Harold, The Measurement of Appearance,(1987)
CIELAB Color Space +b* L* = 116(Y/Y n ) 1/3-16 a* = 500[(X/X n ) 1/3 - (Y/Y n ) 1/3 ] b* = 200[(Y/Y n ) 1/3 - (Z/Z n ) 1/3 ] -a* +a* -b* E* 94 = [( L*/k L S L ) 2 + ( C*/k C S C ) 2 + ( H*/k H S H ) 2 ] 1/2
Color Appearance Model? CIELAB Does: Model Chromatic Adaptation Model Response Compression Include Correlates for Lightness, Chroma, Hue Include Useful Color Difference Measure CIELAB Doesn't: Predict Luminance Dependent Effects Predict Background Surround Effects Have an Accurate Adaptation Transform
Color Reproduction History ART CRAFT??? TECHNOLOGY OPEN SYSTEMS CLOSED SYSTEMS
Example: Color Photography Defined Film Sensitivities and Dyes Controlled Processing Defined Paper Sensitivities & Dyes
Example: Color Television Standard Camera Sensitivities Standard Signal Processing & Encoding Standard Display Decoding and Primaries
Example: Graphic Arts Defined Input Material & Scanner Responses Fixed RGB-to-CMYK Conversion Defined Press, Ink, Paper Characteristics
The End of Closed Systems
Surprise! Expensive Color Devices Didn't Automatically Produce the Desired Results!! "You Need Colorimetric Calibration/Characterization." "But We Know that Won't Work!" "We Need Color Appearance Matching!"
The Problem... The Solution Cross-Media Color Reproduction Colorimetry: Strict Viewing Conditions Color Appearance Modeling: More Realistic Viewing Conditions D93? CRT Display CWF? Room D65 Light Booth D65 CRT Display DEVICE-INDEPENDENT COLOR SPECIFICATION
Hunt's Color Reproduction Objectives 1. Spectral: R(λ) 2. Colorimetric: Yxy 3. Exact: Yxy and Luminance 4. Equivalent: Brightness-Colorfulness 5. Corresponding: Lightness-Chroma 6. Preferred: None of the Above!
Levels of Color Reproduction 1. Color Reproduction 2. Colorimetric Color Reproduction 3. Color Appearance Reproduction 4. Color Preference Reproduction
The Process of D.I.C.I. Input Device Device Independent Color Space Viewing Conditions Independent Space Viewing Conditions Independent Space Device Independent Color Space Output Device Input Device Colorimetric Characterization Chromatic Adaptation & Color Appearance Models Gamut Mapping, Preferences, Spatial Operations, Tone Reproduction, etc. Chromatic Adaptation & Color Appearance Models Output Device Colorimetric Characterization Implementations: ICC, PostScript 2
What's Missing in this Picture? Color Devices Capable OS Device Independent Color Space? Input Profiles CMS Output Profiles
Models, Models, Models! Chromatic Adaptation Models: Corresponding Colors (XYZ) 1 in VC 1 MATCHES (XYZ) 2 in VC 2 If VC 1 VC 2, then (XYZ) 1 (XYZ) 2 Color Appearance Models: Appearance Parameters Lightness, Chroma, Hue Brightness, Colorfulness
Color-Appearance Phenomena Situations in which basic colorimetry fails to specify matches... Chromatic Adaptation Stevens Effect Hunt Effect Helson-Judd Effect Bartleson-Breneman Equations Discounting the Illuminant
Chromatic Adaptation Chromatic Adaptation: Largely independent sensitivity regulation of the (three) mechanisms of color vision 1.25 Relative Cone Responsivity 1 0.75 0.5 0.25 0 400 450 500 550 600 650 700 Wavelength (nm)
1 Stevens Effect 0.1 0.1 1 Y/Yn L (Dark) L (63 db) L (79 db) L (97 db) Stevens & Stevens (1963) Exponent increases with light adaptation. Jameson & Hurvich (1964) Offset increases with light adaptation.
Hunt Effect Corresponding chromaticities across indicated relative changes in luminance (Hypothetical Data) 0.6 0.5 y 0.4 1 10 100 1000 10000 0.3 10000 1000 100 10 1 0.2 0.2 0.3 0.4 0.5 0.6 x Hunt (1952)
Helson-Judd Effect Munsell value and chroma scaling of nonslective samples under a green source on a gray background 8 Hue of Source Chroma 6 4 2 0 Hue of Source's Complement (2) (4) (6) (8) 0 2 4 6 8 10 Value Helson (1938)
Bartleson-Breneman Equations Bartleson and Breneman (1968)
Discounting the Illuminant 0.54 0.52 Adapting Chromaticities Achromatic Chromaticities Incandescent <-- Hands <-- No Hands v' 0.50 0.48 Daylight 0.46 0.18 <-- Hands <-- No Hands 0.20 0.22 u' 0.24 0.26 Fairchild (1992)
Discounting the Illuminant "... in everyday life we are accustomed to thinking of most colors as not changing at all. This is in large part due to the tendency to remember colors rather than to look at them closely." - Evans (1943)
Chromatic Adaptation Models: von Kries "This can be conceived in the sense that the individual components present in the organ of vision are completely independent of one another and each is fatigued or adapted exclusively according to its own function." -von Kries, 1902 L/L max M/M max S/S max
Chromatic Adaptation Models: LLAB Luo et al. (1996) L/L o M/M o (S/S o ) β
Chromatic Adaptation Models: Guth Guth (1995) (L+n) 0.7 [σ/(σ+(l+n) 0.7 )]
Chromatic Adaptation Models: Nayatani Nayatani et al. (1995) [(L+n)/(L o +n)] β
Chromatic Adaptation Models: Hunt Hunt (1994) f n (F L F ρ ρ/ρ w ) + ρ D + 1
Chromatic Adaptation Models: RLAB Fairchild (1996) [p L + D(1-p L )]L/L n
Chromatic Adaptation Models: CIELAB CIE (1986) A "Wrong von Kries" Transform X/X n Y/Y n Z/Z n
Color Appearance Models: CIELAB Munsell Value 5, CIE 1976 a*,b* 80 60 Y L* = 116(Y/Y n ) 1/3-16 b* 40 20 0 G R a* = 500[(X/X n ) 1/3 - (Y/Y n ) 1/3 ] b* = 200[(Y/Y n ) 1/3 - (Z/Z n ) 1/3 ] -20-40 -60-80 B P Fixed Illuminant, Background, & Surround What if we change them? -80-60 -40-20 0 20 a* 40 60 80
Color Appearance Models: Nayatani Input: XYZ stimulus XYZ source Luminance Y background Output: Brightness Lightness Colorfulness Chroma Saturation
Color Appearance Models: Hunt Input: XYZ stimulus XYZ source Luminance XYZ background Y surround Discounting Output: Brightness Lightness Colorfulness Chroma Saturation
Color Appearance Models: RLAB Objectives: Simple Implementation Capture Most-Significant Effects Reference-Condition Corresponding Colors Related Colors (L R, C R, h R ) Image Reproduction M.D. Fairchild and R.S. Berns, Image Color-Appearance Specification Through Extension of CIELAB, Color Res. Appl., 18 178-190(1993). M.D. Fairchild, Refinement of the RLAB Color Space, Color Res. Appl., 21 in press(1996).
RLAB Flow Chart Original Tristimulus Values Original Corresponding Tristimulus Values for Reference Condition Modified CIELAB Values for Reference Condition XYZ X ref Y ref Z ref L R a R b R Original Adapting Luminance & Color, Media Type Original Surround Relative Luminance Relative Appearance Attributes Reproduction Corresponding Tristimulus Values for Reference Condition Reproduction Tristimulus Values L R C R h R X' ref Y' ref Z' ref X' Y' Z' Reproduction Surround Relative Luminance Reproduction Adapting Luminance & Color, Media Type
RLAB Formulation X ref X Y ref = RAM Y Z Z ref M = XYZ-to-LMS R = Reference a L 0 0 A = 0 a M 0 0 0 a S a L = (p L + D(1.0 - p L ))/L n (D = 0.0; No Discounting) (D = 1.0; Discounting) (Intermediate Possible)
RLAB Space L R = 100(Y ref ) σ a R = 430[(X ref ) σ - (Y ref ) σ ] b R = 170[(Y ref ) σ - (Z ref ) σ ] Surround Exponent, σ Average 1/2.3 (0.44) Dim 1/2.9 (0.34) Dark 1/3.5 (0.29) Chroma, Saturation, and Hue determined from L R a R b R
Testing Color Appearance Models in D.I.C.I. Given: Several Color Appearance Models Significantly Different Predictions Requires: Psychophysical Tests Model Performance Model Refinement
Example Images Given a print viewed under incandescent illumination, what image should be produced to match it on a CRT with a D65 white point? Simple Colorimetric Match!
Nayatani Model Match Hunt Model Match CIELAB Match RLAB Match
General RIT Procedure VC 1 VC 2 Original Predictions of 5 Models Viewed Pairwise 2 Interval Scale Rank Order Prediction Quality 1 0-1 -2 RLAB CIELAB von Kries Hunt Nayatani 1. RLAB 2. CIELAB 3. von Kries, Hunt 5. Nayatani Model
Experimental Evaluations 1. Print to Print D65 to A, 3 Luminance Levels, Equal Surround, 4 Images, 30 Observers 2. CRT to Projected Slide D65 & D93 to 3863K, Luminance Change, Surround Change, 3 Images, 15 Observers 3. Print to CRT (Viewing Test) D50 & A to D65, Equal Luminance, Equal Surround, 5 Images, 15 Observers
Experimental Evaluations 4. Print to CRT (Several Conditions) 12 Sets of White-Point, Luminance, Surround, Background 5 Images, 24 Observers 5. Print to CRT (Adjustment) A to D65, Equal Luminance, Equal Surround, 3 Images, 32 Observers
Print-to-Print RANK ORDER: 1. RLAB, CIELAB von Kries, Hunt 5. Nayatani Kim et al. (1993)
CRT-to-Slide RANK ORDER: 1. RLAB 2. CIELAB, von Kries 4. Hunt 5. Nayatani Lester et al. (1995)
Print-to-CRT (Viewing Test) RANK ORDER: 1. RLAB 2. CIELAB 3. von Kries 4. Hunt 5. Nayatani Braun et al. (1996)
Print-to-CRT (Many Conditions) RANK ORDER: 1. RLAB 2. CIELAB 3. von Kries 4. Hunt 5. Nayatani Braun et al. (1996)
Print-to-CRT (Adjustment) RANK ORDER: 1. RLAB, Adjustment, Matrix 3. CIELAB, von Kries 4. Hunt Braun et al. (1996)
CIE TC1-27 Specification of Colour Appearance for Reflective Media and Self-Luminous Display Comparison Chair: Paula J. Alessi, (US, Kodak) Published Guidelines for Coordinated Research Ongoing Experiments RIT, Derby, Chiba, Chinese Culture U.
CIE TC1-34 Testing Colour-Appearance Models Chair: Mark D. Fairchild, (US, RIT) Published Guidelines for Coordinated Research Gathered Data Sets CSAJ, McCann et al., LUTCHI Ongoing Model Evaluations Luo LUTCHI Results Fairchild Other Data Progress Report in Preparation Second Draft: June, 1996 Developing CIE Appearance Model Preliminary Report: May, 1997!
Conclusions Color Appearance Models are Necessary for Device-Independent Color Imaging. Several Models have been Published and Tested. A Simple Model Like RLAB is Adequate for Most Practical Applications. A Complete Model Like Hunt's is Necessary for Specific, Well-Controlled Situations. A Compromise CIE Model Should Include the Best Features of All Models.