Multicomputer system for optical character recognition in web library creation

The described optical character recognition system is based on a multicomputer environment. It performs faster character recognition from printed media and their forwarding to a web library. Software on the server side enables the setting of execution conditions, mapping of pictures with characters to clients and execution monitoring. The developed MOCR mechanism has been implemented on every client. Experimental results enable evaluation of timing performance and recognition accuracy. By increasing the number of machines engaged, makespan is decreased almost linearly. Although more time-demanding, for as many as several engaged machines, MOCR additionally reduces processing time and increases accuracy of recognition.