Constancy of Perceived Depth from Disparity across Spatial Frequency
53.415, Tuesday, 20-May, 8:30 am - 12:30 pm, Banyan Breezeway
Phillip Guan1, Martin Banks1,2; 1Graduate Group in Bioengineering, UC Berkeley and UCSF, 2Vision Science, School of Optometry, UC Berkeley
Contrast constancy is observed in the luminance domain: gratings of different spatial frequencies have equal perceived contrast when they have the same physical contrast despite large differences in contrast thresholds. Seeing depth from disparity is quite different from seeing luminance variation. For example, depth from disparity is constrained by lower and upper bounds: lower-disparity thresholds and upper-disparity limits respectively. We investigated constancy in the disparity domain by comparing the perceived depth of disparity corrugations of different spatial frequencies. We presented a standard stimulus at 0.3cpd and a comparison stimulus at other frequencies. We varied the disparity amplitude of the comparison to determine the value yielding the same perceived depth as the standard. When the standard was near its lower-disparity threshold, more disparity was required in the comparison for equal perceived depth. When it was above its lower threshold, equal perceived depth occurred when the comparison had the same disparity; i.e., constancy was observed. When the disparity of the standard approached the upper-disparity limit, its perceived depth dropped precipitously. Equal perceived depth then occurred when the comparison’s disparity was less than the standard’s. Near the upper-limit the task could not be done with frequencies higher than the standard because these stimuli exceeded the disparity-gradient limit. Thus, depth constancy occurs across a broad range of corrugation frequencies provided that the disparity amplitude is not close to the lower-disparity threshold or upper-disparity limit. Low spatial frequencies avoid the disparity-gradient limit at all but the greatest amplitudes and can convey more apparent depth than other frequencies. Our findings call into question stereo compression algorithms that use lower-disparity thresholds to manipulate disparities in images and video and show that disparities near the upper-disparity limit warrant special consideration when creating stereo content.