hide
Free keywords:
-
Abstract:
We construct a model of intracortical interactions in V2 to capture the following physiological and psychophysical phenomena: 1) depth edge highlighting, a V2 cell often responds more vigorously when its receptive field (RF) is near a depth edge (von der Heydt et al 2000); 2) disparity capture, V2 responses are tuned to the depth values at surface boundaries beyond their RFs (Bakin et al 2000); 3) popout, a target pops out perceptually from a background of distractors at a different depth; and 4) transparency, two superposed random dot stereo depth planes perceptually segregate. Each model cell is binocular and tuned to a specific depth. The input stimuli (and RFs) are individual dots in visual space without orientation, color, or motion information. The model inputs and initial responses include both correct and false matches between stimuli dots in the two eyes, as suggested by the responses of V1 cells (Cumming and Parker, 2000), which provide inputs to V2. Finite range horizontal connections link model cells tuned to similar depth values. Nearby model pyramidal cells influence each other via mono-synaptic facilitation or disynaptic inhibition. Contextual influences modify the cell responses, such that activities in the model are ultimately significant only from cells whose preferred disparities correspond to the correct matches. This models disparity capture and transparency. Furthermore, responses are higher to the lone target dot of a unique disparity, or dots near a depth edge, enabling pop-out and region segmentation. We suggest that the contextual influences determine many stereo grouping phenomena.
