Mechanisms of Color Detection, Induction and Adaptation

Possibly the most important functions of color vision are the identification of objects and materials across illuminants, and of illuminants across sets of objects. In the color constancy literature, almost all experimental work, and most computational models, suggest that the nervous system should discount the illuminant so that colors of objects appear constant. This project, however, is based on the contrasting observation that, in many situations, objects do appear to be of systematically different colors under different illuminants. When the illuminant spectrum changes, changes in the spectra of reflected lights result in similar multiplicative changes in cone absorptions for all natural objects. Using real and simulated stimuli, we will ask whether observers can use these shifts to identify objects and illuminants, using similarity of appearance or comparisons of shifts in appearance. Where identification across illuminants depends on comparing color changes in objects to color changes in backgrounds, we will derive the geometric color relations that are used. We will study the intrinsic geometry of perceptual color space by using theorems as consistency checks, and test its invariance to illuminant changes. Most objects in the world are not uniform in color, so we will test patterned objects to see whether identification of local contrasts helps in identifying materials. When only a single illuminant is present, constancy of appearance is the best cue for material identification. We will perform tests of whether constancy mechanisms extend over space or are spatially local but extend across time, and also test whether the results require models of adaptation or perceptual frames of reference. We will directly measure the perceived color of the illuminant, identify the scene statistics that are used in this estimation, and test if this estimate is used in color constancy. The results of this project will provide insights into neural computations and representations underlying everyday color tasks. The methods developed here will be useful in identifying functional consequences of color vision deficits. PHS 398 (Rev. 05/01) Page 2 Principal Investigator/Program Director (Last, first, middle): ZAIDI, QASIM