A parallel architecture for performing real time multi-line optical character recognition

The design of a multiprocessor parallel architecture for performing real-time multiline optical character recognition (MLOCR) is described. MLOCR is a computationally intensive application involving real-time digitization of images, pattern recognition, feature extraction, scaling, rotation and gray-scale thresholding, and character recognition. System organization, control and subsystem partitioning of MLOCR algorithms is discussed. System interprocess communications protocols and control of the multiple asynchronous processors are covered. Functional partitioning of algorithms onto each processor, bus definitions based on real time data communications bandwidths, and system control requirements are also discussed.