Stereo Matching by Hierarchical, Microcanonical Annealing

Abstract

An improved stochastic stereo-matching algorithm is presented. It incorporates two substantial modifications to an earlier version: a new variation of simulated annealing that is faster, simpler, and more controllable than the conventional "heat-bath" version, and a hierarchical, coarse-to- fine-resolution control structure. The Hamiltonian used in the original model is minimized, but far more efficiently. The basis of micro canonical annealing is the Creutz algorithm. Unlike its counterpart, the familiar Metropolis algorithm, the Creutz algorithm simulates a thermally isolated system at equilibrium. The hierarchical control structure, together with a Brownian state-transition function, tracks ground states across scale, beginning with small, coarsely coded levels. Results are shown for a 512 x 512 pair with 50 pixels of disparity.