Role of chromatic and luminance contrast in inferring structure from motion.

Abstract

We measured the ability to infer structure from motion (SFM) in several directions in three-dimensional color space. Only motion cues are useful to subjects in performing this three-dimensional shape-identification task. We report the following results: (1) SFM performance is at chance for equiluminant stimuli that isolate short-wavelength-sensitive cones. Hence the short-wavelength-sensitive-cone input to SFM is negligible. (2) SFM performance increases with the magnitude of delta L - delta M signal when delta L + delta M = 0 (i.e., only chromatic and no luminance contrast is available). We reject the hypothesis that SFM obtains input from a single chromatic mechanism combining the long- and medium-wavelength-sensitive cones linearly. Our data are compatible with SFM that uses the output of two mechanisms, one taking the difference between the long- and medium-wavelength-sensitive-cone signals and the other taking the respective sum. We reject the particular hypothesis that SFM utilizes only the magnitude and not the sign of the long- and medium-wavelength-sensitive-cone signal. (3) We compare SFM performance with threshold performance for velocity and motion discrimination. Stimuli with luminance contrast yield SFM performance that is superior to stimuli without luminance contrast when they are expressed as multiples of velocity discrimination threshold. This superiority is even greater when SFM performance is compared with motion-direction discrimination thresholds.