Color Appearance Models

• Should account for chromatic adaptation and illumination.

• CIELab does account for illumination and can be modified to account for chromatic adaptation by scaling cone responses independently after converting XYZ to cone coordinates. But:

Color Appearance ModelsProblems with CIELab

• CIELab is not a simple adaptation model: [Xa,Ya,Za]t = M-1 diag(kL kM kS) M [X Y Z] t

• is not itself a linear adaptation model, i.e.is not of the form

[Xa,Ya,Za]t = diag(kL0 kM0

kS0) [X Y Z] t

• (“A wrong Von-Kries transformation”

Color Appearance ModelsProblems with CIELab

• Trouble for Munsell colors:– Constant Munsell chroma and hue are not

constant in a* b* space (Fairchild Fig 10-4)– Munsell R, G, Y, B do not lie on a*,b* axes

• No prediction of luminence dependencies

Nayatani Color Appearance Model

• Foundation in illumination engineering– need to model illumination color rendering– how general?

• Models appearance of uniform patches on gray background

• Not good for complex stimuli or backgrounds

Nayatani Color Appearance Model

• Data for the model– Background reflectance Y0 > 18%

– CIE chromaticity x0 y0 of illuminant

– CIE chromaticity x y and reflectance Y of stimulus

– Absolute illuminance E0 of viewing field in lux• http://www.intl-light.com/handbook/irrad.html

Nayatani Color Appearance Model

• Model components– Nonlinear chromatic adaptation– One achromatic, two chromatic color opponent

channels weighted by cone population ratios

Nayatani Color Appearance Model

• Model outputs– Brightness as linear function of adapted cone

responses (which are non-linear!)– Lighness: achromatic channel origin translated

to black=0, white = 100– Brightness of “ideal white” (=perfect reflector)– Hue angle (from the chromatic channels)

Nayatani Color Appearance Model

• Model outputs– Hue quadrature: interpolation between 4 hues defined

by chromatic channels red (20.14), yellow (90 .00), green (164.25), blue (231.00)

– Saturation: depends on hue and luminance (predicts changes of chromaticity with luminance)

– Chroma = saturation*lightness

– Colorfullness: Chroma*brightness of ideal white.

Nayatani Color Appearance Model Advantages

• Invertible for many outputs, i.e. measure output quantities, predict inputs

• Accounts for changes in color appearance with chromatic adaptation and luminance

Nayatani Color Appearance Model Weaknesses

• Doesn’t predict:– Effects of changes in background color or relative luminance

– incomplete chromatic adaptation– cognitive discounting the illuminant– appearance of complex patches or background– mesopic color vision