A New Hopfield Neural Network-Based Stereo Matching

Abstract

A Hopfield neural network-based sterro matching algorithm is presented in this paper. We formulate an effective energy function which is combined with correlation, uniqueness, epipolar line, disparity and 0-1 integer properties. This energy function is minimized by a two-dimensional asynchronous Hopfield neural network. This proposed method is implemented and compared with Nasrabadi’s approach. It is found that proposed method is superior in computation speed and feature matching performances.