Optimizing a video preprocessor for OCR

This paper describes how optimal video preprocessor performance can be achieved using a software recognition system and a set of controlled experiments. It basically involves optimizing on a constant threshold and then using recognition logic designed from that threshold to optimize an adaptive threshold function or threshold operator. By using this technique, the video preprocessor can be optimized early In the development of an OCR (Optical Character Recognition) machine. The software recognition system is described and experimental data is presented to illustrate the procedures.