Disparity capture by flanking stimuli: A measure for the cooperative mechanism of stereopsis

The cooperative algorithm first proposed by Marr and Poggio (1976) solves correspondence problem in stereo matching via long-range excitatory horizontal connections between neurons tuned to nearby disparity values. This algorithm served as a basis for other models of stereopsis, including Prazdny's (Prazdny, Bio. Cyber. 1985) and PMF (Pollard, Percep.,1985) matching algorithms. An extensive range of excitatory and inhibitory connections found in both V1 and V2, and contextual effects of stimuli beyond the classical RF provide physiological support for the algorithm. In this work we measure the range and scaling properties of the cooperative interaction by studying the effect that flanking Gabor patches produce on a perception of a target pair of patches. The relative depth configuration of the target pair can switch from the small disparity gradient to a large disparity gradient state (originally a “false” match) as a result of contextual effects of the flanking stimuli. We found that the interaction strength falls exponentially with distance. It's range varies for different subjects from 2 to 3 Gabor patch sizes and scales proportionally to the size of the stimuli. The results suggest that a localized cooperative interaction is in effect at a broad range of spatial scales