Friday Lunch Seminar (English)June 3, 2022
12:15 〜 13:00
(On-line)

Talk Title: Cross-correlation and cross-matching binocular disparity representations across the human visual cortex

Bayu Gautama Wundari
Researcher
Brain Networks and Communication Laboratory

Host PI: Hiroshi Ban

Abstract:
Perceiving binocular disparity-defined depth in visual scenes requires identifying the corresponding points in the left and right retinal images. The initial stage of solving this stereo correspondence problem occurs in V1, representing binocular disparity cues based on a computation similar to the cross-correlation. The represented information still carries binocularly mismatched signals and therefore needs to be transformed into a cross-matching representation in which the mismatched responses are eliminated. It remains elusive where the transformation from the cross-correlation into the cross-matching representation occurs across the human visual cortex. We engineered random-dot stereograms in three dot-match levels: anti-correlated, half-matched, and correlated dots. We fitted the brain region’s representation elicited by those stimuli with the linear weighted sum of the cross-correlation and the cross-matching representational model outputs. The model’s weights suggest that the response patterns in the studied brain regions comprised these two representations with different proportions. The transition from the cross-correlation into the cross-matching occurred after V3. Our result suggests that the brain solves the stereo-correspondence problem by transforming the cross-correlation representation in the early cortex into the cross-matching representation in the higher cortex.