This report examines a number of parallel algorithms for solving random-dot stereograms. A new class of algorithms based on the Boltzmann Machine is introduced and compared to previously developed algorithms. The report includes a review of the stereo correspondence problem and of cooperative techniques for solving this problem. The use of energy functions for characterizing the computational problem, and the use of stochastic optimization techniques for solving the problem are explained.
|Institution||Computer Science Department, Carnegie Mellon University|