Statistical mechanics of stereoscopic vision

Publisher Website
Google Scholar
Abstract
A crucial issue in both mammalian and machine vision is how depth is perceived. In this paper we explore the relationships between the optimization tasks involved in stereoscopic vision and statistical mechanics. The ability of different algorithms in processing visual information to find stereoscopic disparity is examined. An algorithm to perform this task is introduced by expressing the optimization constraints in terms of a generalized spin-glass Hamiltonian. Our main results concern the derivation of the exact free-energy surface of a global version of the model in the spirit of the Sherrington-Kirkpatrick approach to spin glasses.

This publication has 39 references indexed in Scilit: