Abstract
Human vision strives to make the perceived color of objects relatively unchanged under shifts in illumination – a capacity termed "color constancy". To discount the chromaticity and intensity of illumination the visual system must estimate it from light reflected from surfaces in areas surrounding a target object, and from past stimulation. These computations are done by neural mechanisms of spatial color contrast and temporal adaptation. Temporal adaptation may cause a visible afterimage, the hue of which is approximately complementary to that of the adapting stimulus. Similarly, color contrast may shift color in a direction that is approximately complementary to its surround. We probe cone excitation space, by modulating stimuli along the ±(S-cone) and ±(L-M) as well as other chromatic axes, to characterize low level mechanisms involved in discounting illumination. We use asymmetric color matching to measure the chromaticity produced by spatial induction and of afterimages. Results show that neither afterimages (Loomis, 1972) nor spatially induced colors are actually complementary to the colors of the inducing stimuli. The important exceptions are ±(L-M) stimuli, which do produce complementary afterimages and complementary spatial chromatic induction. In all cases, the match to the color that is spatially induced has the same relative S/(L-M) difference from complementarity as the color that matches the afterimage, indicating that afterimages and spatially induced color have the same hue. That this hue is complementary only for stimuli that do not modulate S-cones strongly suggests an interaction of cone systems that can only be explained by a subtractive postretinal process, one that requires S cone contributions to multiple pathways -- contradicting the cardinal axis model. The data show that these S cone contributions are relatively the same for both afterimages and spatial chromatic induction and lead to a model framework for the mechanisms that discount changes in illumination.
Meeting abstract presented at VSS 2016