Color constancy and color discrimination

Abstract

I have been studying the implications of the hypothesis that the mechanisms of light adaptation compute an internal representation of the retinal image that discounts the spectral power distribution of the ambient light in order to estimate the surface spectral reflectances of objects in the image. I review two consequences of this hypothesis for color discrimination. First, I describe the formulas for estimating surface spectral reflectances from the photoreceptor quantum absorptions. From these formulas, I derive predictions for color discrimination based on the estimated surface spectral reflectances. The structure of the color discrimination equations based on surface spectral reflectance estimates is essentially the same as MacAdam and Silberstein’s line-element equations. The interpretation of the equations with respect to empirical measures—such as wavelength discrimination and the MacAdam ellipses—is somewhat different. The most important difference arises because the observer’s estimate of the local ambient light spectral power distribution, rather than a central measure of the local quantum catch, is assumed to govern the state of light adaptation. Second, I describe how we are attempting to use the color difference measurement based on estimated surface reflectance to predict color differences in spatially complex displays.