Abstract: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.

Structure from stereo by associative learning of the constraints

Modeling pre-attentive stereo grouping by intracortical interactions in early visual cortex

From Binocular Input Correlation to Disparity Tuning and Ocular Dominance in Cortex

Efficient Stereo Coding in the Multiscale Representation

Pre–attentive Segmentation and Correspondence in Stereo

Pre-attentive segmentation and correspondence in stereo.

Incorporating Visual Knowledge Representation in Stereo Reconstruction

Learning to Adapt for Stereo

Neural architectures for stereo vision

Robust stereo matching with surface normal prediction.

Stacking Learning with Coalesced Cost Filtering for Accurate Stereo Matching

Modeling stereopsis via Markov random field

Evidence of Stereoscopic Surface Disambiguation in the Responses of V1 Neurons

Stereo Matching Via Disparity Estimation And Surface Modeling

Polka Lines: Learning Structured Illumination and Reconstruction for Active Stereo

Propagation of Uncertainty for an Epipole-Dependent Model for Convergent Stereovision Structure Computation

Learning Structure Models with Context Information for Visual Tracking.

Effects of image statistics on stereo coding in human vision

A stereo advantage in generalizing over changes in viewpoint on object recognition tasks

Connecting The Dots: Learning Representations For Active Monocular Depth Estimation

Learning intermediate-level representations of form and motion from natural movies