A Bayesian approach to the stereo correspondence problem

Authors:
Jenny C. A. Read
Affiliations:
University Laboratory of Physiology, Oxford, OX1 3PT, U.K.
Venue:
Neural Computation
Year:
2002

Citing 7
Cited 6

Bayesian modeling of uncertainty in low-level vision

International Journal of Computer Vision
Perception as Bayesian inference

Perception as Bayesian inference
Binocular receptive field models, disparity tuning, and characteristic disparity

Neural Computation
Stereo Matching with Nonlinear Diffusion

International Journal of Computer Vision
A Deterministic Approach for Stereo Disparity Calculation

ECCV '92 Proceedings of the Second European Conference on Computer Vision
A Computational Framework for Determining Stereo Correspondence from a Set of Linear Spatial Filters

ECCV '92 Proceedings of the Second European Conference on Computer Vision
Occlusions and Binocular Stereo

ECCV '92 Proceedings of the Second European Conference on Computer Vision

Normalization enables robust validation of disparity estimates from neural populations

Neural Computation
Orientation disparity: A cue for 3d orientation?

Neural Computation
Disparity estimation by pooling evidence from energy neurons

IEEE Transactions on Neural Networks
Autonomous development of vergence control driven by disparity energy neuron populations

Neural Computation
Modeling stereopsis via markov random field

Neural Computation
2013 Special Issue: Event-based 3D reconstruction from neuromorphic retinas

Neural Networks

Quantified Score

Hi-index	0.01

Visualization

Abstract

I present a probabilistic approach to the stereo correspondence problem. Rather than trying to find a single solution in which each point in the left retina is assigned a partner in the right retina, all possible matches are considered simultaneously and assigned a probability of being correct. This approach is particularly suitable for stimuli where it is inappropriate to seek a unique partner for each retinal position-for instance, where objects occlude each other, as in Panum's limiting case. The probability assigned to each match is based on a Bayesian analysis previously developed to explain psychophysical data (Read, 2002). This provides a convenient way to incorporate constraints that enable the ill-posed correspondence problem to be solved. The resulting model behaves plausibly for a variety of different stimuli.