Purchase this article with an account.
Zhaoping Li; Modeling pre-attentive stereo grouping by intracortical interactions in early visual cortex. Journal of Vision 2001;1(3):168. doi: https://doi.org/10.1167/1.3.168.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Background: The responses of disparity selective cells in V2 can be influenced by stimuli outside their receptive fields. In particular, the response is often higher when the receptive field is near a depth edge, e.g., a border between two different depth planes composed of random dots (von der Heydt, Zhou, and Friedman, Vision Research, 2000). In addition, cell responses can reflect the phenomenon of disparity capture, in which global surfaces or grouping cues dictate the local interpretation of stimulus depth. Hence, cell responses to the surface area of a depth plane are tuned to the depth values at the surface boundary beyond the receptive fields (Bakin, Nakayama, and Gilbert, J. Neuroscience, 2000). Method: A model of the cortex with intra-cortical interactions is constructed to study how contextual influences contribute to the stereo grouping and segmentation phenomena observed physiologically. Nearby model pyramidal cells influence each other via monosynaptic facilitation or disynaptic inhibition. Input and neural representation focuses on the stereo domain. Hence, the input stimuli are dots in depth space without orientation, color, or motion information. The model cells are binocular and receive inputs from both the correct and false matches between stimuli dots in the left and right eyes, as suggested from the physiological data about the responses of V1 cells (Cumming and Parker, J. Neuroscience, 2000), which provide inputs to V2. Hence, the initial responses to an input pattern come from many model cells corresponding to all possible stereo matches. Contextual influences modify the cell responses, such that after the initial transient, the responses from the model are significant only from cells whose receptive fields correspond to the correct matches only. In particular, the model reproduces the physiological disparity capture and depth edge highlights.
This PDF is available to Subscribers Only