Farshad Moradi, School of Intelligent Systems, Institute for Studies in Theoretical Physics and Mathematics, Tehran, Iran
Submitted abstract to 1999
AAO annual meeting
It's shown that complex binocular cells in primary visual cortex can act as depth estimators. However, interaction between these estimators is essential for recovering phase singularities and object transparencies. In a previous study, I had proposed that depth estimators are arranged in functional layers according to their preferred disparity. Here interactions between different layers are investigated.
Single image random dot stereograms with equalized r.m.s. were used as stimuli. In a 8-alternative forced choice task, subjects were asked to fade out some fixed mask and fade in a target (a circle) until they could accurately discriminate the target and it's position. The contrast of target was measured as discrimination threshold. In some trials, ideal high pass (>8cpd) and low pass (<4cps) filters (one octave apart) were used for target and mask, respectively. Diameter of target was 2º, and it was placed 3º from fixation point.
Threshold with a blank mask (disprity=0') depends on stimulus disparity (ANOVA, P<.004. 4' vs. 8' and 12': P=.005,.01. 8' vs. 12': NS). For unfiltered trials with positive target and mask disparity, thresholds are lower when mask is behind the target (target/mask disparity: 8'/4' vs. 8'/12' : P<.002, 8'/12' vs. 12'/8': P=.000001). However for filtered stimuli thresholds are higher and asymmetry is reversed. (8'/12' vs. 12'/8': P=.014 for additive transparency, P<.0002 for second order transparency). Depth asymmetry is also reversed when the sign of target's disparity is the opposite of mask's (4'/-4' vs. -4'/4': P<.000001).
The differences between thresholds suggest complex interactions between cells tuned for different disparities. Even in quite similar tasks, spatial frequency content or disparity itself can affect the process of binocular matching. I propose that these asymmetries are computationally meaningful, and reflect probabilities of the real world.
