Abstract
Binocular disparity is a precise depth cue and, in view of its derivation from binocular geometry, is generally considered to be independent of luminance contrast. Here, we provide evidence showing otherwise. The test stimuli were rectangular random dot stereograms (1.27 x 3.44 degree) of 100% density light and dark dots of 1 arc min extent that produced a percept of a surface with a sinusoidal modulation (0.29 or 0.87 cycle/deg) in depth. The observers' task was to adjust the length of a horizontal bar to match the perceived depth in the test stimuli as a function of the contrast of the random dots. The luminance contrast of random-dot array ranged from 5% to 80% of the mean luminance. At each contrast level, the matched depth increased with disparity up to about 10 arc min and then decreased with further increases in disparity. Both the amplitude and the peak position of the matched depth function increased as a sigmoid function of luminance contrast, only leveling out by about 40% contrast. These results show that perceived depth from disparity modulation depends strongly on the luminance contrast of the image. This result cannot be explained by a disparity energy model or any model that postulates the perceived depth as being determined by the disparity channel that gives the greatest response to stimuli, because those models would predict no effect of luminance contrast on perceived depth. Instead, the data can be explained by a multiple-stage model in which the perceived depth is determined by the weighted average of several nonlinear contrast gain control mechanisms, each with a different disparity selectivity.
Meeting abstract presented at VSS 2016