Spectral Adaptation Mark D. Fairchild RIT Munsell Color Science Laboratory IS&T/SID 14th Color Imaging Conference Scottsdale 2006 Chromatic Adaptation
Spectra-to-XYZ-to-LMS Chromatic adaptation models are viewingcondition-dependent transformations of LMS. 1 Relative sensitivity 0.75 0.5 0.25 S M L 0 380 480 580 680 780 Wavelength, nm
What if you want spectra on the output end? LMS-to-XYZ-to-Spectra Undefined One-to-Many Flashback to CIC13
Noise Adaptation Noise Adaptation
Noise Adaptation Channel-Free Model CSF Original Image FFT of Image CSF CSF a = FFT(im) +1 Spatial Filtered Image Smoothed FFT Can it work with color spectra too???
The Model Illuminant/Source WL (nm) Illuminant/Source Reflectance WL (nm) Reflectance! Define Blur! Stimulus Equal-Energy Illuminant (1-D) Blurred Illuminant/Source + (D) Adapting Stimulus Adapted Stimulus Adapted Ill. E Reflectance WL (nm) Colorimetry Illuminant E XYZ CIELAB Step-by-Step Illuminant/Source WL (nm) Reflectance WL (nm) Illuminant/Source Reflectance!
Define Blur Step-by-Step!! Stimulus Equal-Energy Illuminant (1-D) Blurred Illuminant/Source + (D) Adapting Stimulus Adapting Step-by-Step Stimulus Adapted Stimulus Adapted Ill. E Reflectance WL (nm) Colorimetry Illuminant E XYZ CIELAB
The Model Illuminant/Source WL (nm) Illuminant/Source Reflectance WL (nm) Reflectance! Convert Stimulus & Adapting Stimulus to Wavenumber Define Blur! Stimulus Spectrally Blur Adapting Stimulus Equal-Energy Illuminant (1-D) Blurred Illuminant/Source + (D) Blend Adapting Stimulus with Ill. E Adapting Stimulus Divide Stimulus by Blurred & Blended Adapting Stimulus Adapted Stimulus Convert Back to Wavelength and Compute Colorimetry Adapted Ill. E Reflectance WL (nm) Colorimetry Illuminant E XYZ CIELAB Color Constancy No Blur of Adapting Spectrum D=1 (Complete Adaptation) Amounts to colorimetry on reflectance curves with no illuminant.
About that Wavenumber scale... Cones on Wavelength 2.000 1.800 1.600 Cone Responses 1.400 1.200 1.000 0.800 0.600 2.000 L M S Cones on Wavenumber 0.400 0.200 1.800 0.000 1.600 400.000 450.000 500.000 550.000 600.000 650.000 700.000 Wavelength 1.400 H.J.A.Dartnall, The interpretation of spectral sensitivity curves, Brit. Med. Bull. 9, 24-30 (1953). Cone Responses 1.200 1.000 0.800 0.600 0.400 0.200 0.000 14000 16000 18000 20000 22000 24000 26000 L M S Wavenumber (cm-1) A Quick Experiment Appearance Scaling (Lightness, Chroma, Hue) for One Observer GretagMacbeth ColorChecker Chart (24 Patches) GretagMacbeth Spectralight III Booth (A, D75, TL84, CWF, Hor)
Results Median CIELAB Color Differences Various Models vs. Observed Various Models vs. CIECAM02 Models: CAT02, CIELAB, Spectral, Constancy Compared with Data 40.0 Median CIELAB Color Difference 35.0 30.0 25.0 20.0 15.0 10.0 5.0 A D75 TL84 Hor CWF 0.0 Spectral CAT02 CIELAB Constancy Adaptation Model
Computational Comparison CAT02 Reference Other Models Illuminant/Source Reflectance Illuminant/Source Reflectance!! Stimulus Stimulus CAT02 Spectral, CIELAB, or Constancy Corresponding Ill. E Colorimetry Corresponding Ill. E Colorimetry " CIELAB Color Differences Compared with CIECAM02 14.0 Median CIELAB Color Difference 12.0 10.0 8.0 6.0 4.0 2.0 A D75 TL84 Hor CWF 0.0 Spectral CAT02 CIELAB Constancy Adaptation Model
Physiological Plausibility? Multiple CMFs 2 1.5 Tristimulus values 1 0.5 0 380 480 580 680 780 Wavelength, nm
Yesterday s Poster... C. Liu and M.D. Fairchild, Color matching a display and its surround, IS&T/SID 14th Color Imaging Conference, Scottsdale, in press (2006). Model Before After Imai et al. F. H. Imai, R. S. Berns and D. Tzeng, A comparative analysis of spectral reflectance estimation in various spaces using a trichromatic camera system, J. Imaging Sci. Technol. 44, 280-287 (2000).
Color Filter Array (RGB) and Two Absorption Filters Filter 1 (RGB)filter 1 Signal Processing Bayer Pattern Sensor Filter 2 (RGB)filter 2 6 Channels Reflectance Factor Wavelength (380-730 nm) Similarity: Macula & 6- Channel Filter... 1.5 Lens Macula Optical Density 1.0 0.5 0.0 400 500 600 Wavelength (nm) 700
Inter-Reflections Objects Under White Light
Chromatic Light With Inter-Reflections
3 Patches of Colored Paper 3 Filters to Illuminate Them
The Results Blue Patch / Red Light Purple Patch / Green Light Yellow Patch / Blue Light The Results Blue Patch / Red Light Purple Patch / Green Light Yellow Patch / Blue Light
Cubes Cubes with Filtered Light Blue Cube / Red Light Purple Cube / Green Light Yellow Cube / Blue Light
Cubes with Filtered Light Blue Cube / Red Light Purple Cube / Green Light Yellow Cube / Blue Light Cubes with Filtered Light Blue Cube / Red Light Purple Cube / Green Light Yellow Cube / Blue Light
Adaptation to Full Scene Inter-Reflections Reveal Spectral Information About the Objects and Help Separate Source & Object Influence
Conclusions A spectral adaptation model can perform similarly with chromatic adaptation models. Such a process could be useful in spectral imaging work-flows. The visual system does have access to more than 3 dimensions of color information in real-world viewing situations. Thank You...