Author(s)

Dimitrios Tzovaras and Michael G. Strintzis

Abstract

A neural network based algorithm is presented for solving the stereo vision corre- spondence problem. The stereo images are divided into blocks and for each block in the left image its corresponding one in the right image is found. The problem is presented as the minimization of a cost function which can be the Lyapunov function of a two-dimensional binary Hopeld neural network. The states of the neurons are updated so as to minimize the cost function. The updating procedure is iterated until the network settles to a stable state. After running the network some of the matched blocks have multiple matches. A post processing procedure is used f

Keywords