Doi, T., Takano, M., Fujita, I.
Binocular correlation and matching computations distinctly contribute to depth perception for rapid and slow stimuli.
Stereoscopic depth perception depends on differences between left and right retinal images (binocular disparity). The primate brain encodes binocular disparity by using the two major visual pathways. The dorsal pathway computes disparity by correlating binocular images (correlation computation). The ventral pathway computes disparity using binocularly matched features (matching computation) such that it removes disparity signals from stereograms without binocularly matched patterns (e.g., contrast-reversed random-dot stereograms). Here we examined the two disparity computation processes in depth perception using slowly and rapidly changing stimuli. To manipulate stimulus speed, we used dynamic random-dot stereograms (DRDSs) and renewed their dot patterns at various refresh rates. The stimulus strengths of the DRDSs were varied by reversing the luminance contrast of dots in one eye at varying proportions. If the contrast of all dots is reversed, only correlation computation encodes disparity because such DRDSs contain no binocularly matched features but retain negative binocular correlation (no-match DRDSs). If the contrast of half dots is reversed, only matching computation encodes disparity because of the zero overall correlation (no-correlation DRDSs). Human subjects were asked to report whether the center region of a DRDS was perceived nearer or farther than its adjacent surround. In each block of trials, we fixed the pattern refresh rate of the DRDSs but varied the disparity sign and the proportion of contrast-reversed dots in the DRDSs’ center region. For totally contrast-reversed (no-match) DRDSs, subjects perceived reversed depth at a fast refresh rate (42.5 Hz). The reversed depth perception weakened with decreasing refresh rates. For half-contrast-reversed (no-correlation) DRDSs, subjects correctly perceived depth at a slow refresh rate (5.3 Hz). The correct depth perception weakened with increasing refresh rates. We modeled the depth discrimination at varying refresh rates using weighted averages from correlation and matching computations. The relative weight for matching computation decreased with increasing refresh rates. From these results, we conclude that stereoscopic depth perception distinctly reflects correlation-based and match-based disparity computations according to the stimulus refresh rate.