Modelling biological depth perception in binocular vision: the local disparity estimation.

Abstract

This paper presents an approach to solving the correspondence problem in binocular vision and to computing the local horizontal disparity map using a biologically inspired algorithm. A computer application was developed as a tool for implementing, developing, and testing computational models for stereopsis, and also as a framework for integrating the disparity map with other perspective clues. Two models for stereopsis have been implemented. One of them is biologically inspired (it models the behaviour of simple and complex cells from the striate cortex) and the other is the 'classical' model of David Marr and Tomaso Poggio, implemented in order to have a comparison term for the simulation results. The paper details the results obtained on random-dot stereograms and on pairs of real images.