Parallel Architectures and Algorithms for Image Understanding

Several issues in using parallel processing for image understanding are addressed. First, efficient schemes for parallel image array access are developed. New class of latin squares called perfect latin squares are defined. Several construction methods are shown and some useful properties of such squares are identified. A generic parallel model of computation employing electro optical devices is developed. Parallel techniques for image computations are developed on this model. Finally, processor time optimal solutions to several low and intermediate level image understanding tasks are designed on orthogonal access parallel architectures.