Abstract
Previous studies of motion parallax have employed random dot textures, which are broadband in terms of spatial frequency and orientation. However most neurons in the early visual system have specific tuning for spatial frequency and orientation. Furthermore, neurons selective for texture boundaries exhibit distinct tuning for high spatial frequency textures. Here we examine the effect of texture spatial frequency and orientation on depth perception from shear motion parallax. Visual stimuli consisted of textures created from randomly distributed Gabor micropatterns whose relative shearing motion was synchronized to the observer's horizontal head movements and modulated with a low spatial frequency (0.1 cpd), horizontal square wave envelope pattern. We measured psychophysical performance in a 2AFC depth-ordering task, for Gabor elements of varying spatial frequency (1 to 8 cpd) and orientation (vertical or horizontal). All of the Gabor micropatterns in each texture were of the same spatial frequency and orientation. Performance was measured for varying levels of added coherence noise, to obtain coherence noise thresholds. Furthermore, we varied the density and the contrast of Gabor micropatterns to measure the possible importance of sparseness and element contrast. At low spatial frequencies, performance was better for vertical than for horizontal Gabors while at high spatial frequencies (e.g. 8 cpd) there was no effect of orientation. However at mid-range spatial frequencies (e.g. 4 cpd), surprisingly, depth for most observers was better for horizontal than for vertical Gabors. Density of the micropatterns had little impact on psychophysical performance. Decrease in contrast increased the difference between performance for vertical and horizontal Gabor micropatterns. These results demonstrate that the mechanism for depth from motion parallax is highly dependent on the nature of the constituent surface textures.
Meeting abstract presented at VSS 2014