Purchase this article with an account.
Laurence T. Maloney, Katja Doerschner, David H. Brainard; Color constancy in 3D scenes: contrasting illumination-estimation and heuristic models. Journal of Vision 2007;7(9):458. https://doi.org/10.1167/7.9.458.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: The illumination of a surface patch can change with the location or orientation of the patch. Recent work indicates that observers partially discount the illumination in judging the color and lightness of matte surfaces. Whether this constancy is most usefully understood in the context of illumination-estimation models or as resulting from the action of a set of heuristics is a matter of current dialog. Here we test whether a coplanar ratio heuristic (see Gilchrist 1977) can parsimoniously account for performance. The test is whether the judged color of a surface depends strongly on the chromatic properties of light reflected from other scene surfaces that are coplanar with it.
Methods: Stimuli were computer-rendered stereo scenes, illuminated by a yellow punctate source and a blue ambient source. Embedded in each scene were two intersecting arrays of coplanar patches, one oriented at 45° azimuth and the other at −45°. One array (B) contained bluish surfaces while the other (Y) contained yellowish surfaces. A test patch near the center of the scene was presented at 45°, 0° or −45° azimuth. Observers adjusted the test patch until it appeared achromatic.
We manipulated the simulated reflectance of scene surfaces to dissociate the effects of coplanar surfaces from those of the scene illumination. In our initial experiments, the manipulation was a spatial reversal of the embedded B and Y arrays. We studied the effect of this manipulation for both sparse and rich scenes.
Results: The B–Y reversal of the coplanar arrays had no discernable effect on the achromatic settings. This result casts doubt on the general utility of the coplanar ratio principle. Further experiments will attempt to characterize the stimulus domain where this heuristic applies, and compare this to the domain where illumination-estimation models are effective.
This PDF is available to Subscribers Only