Abstract

Recent work in computational color vision¹,² demonstrates that it is possible to achieve exact color constancy using only the kinds of information available to biological vision systems. These algorithms first compute an estimate of the illuminant in the scene and then transform the initial color signal so as to eliminate the contribution of the illuminant. We examine this work and relate it to research in color vision through the following linking hypothesis: local retinal state is chosen to compensate for the illuminant. With this linking assumption, we demonstrate that the computational work speaks directly to current and past models of adaptation in visual color psychophysics. Further, the work emphasizes the distinction first clearly drawn by Stiles between the local adaptation state of the retina and the (global) control of adaptation and suggests a broad class of experiments addressing these issues.

© 1986 Optical Society of America