Multilingual character segmentation using matching rate

Publisher Summary Character segmentation which has an effect on the performance of optical character recognition (OCR) system is very difficult especially when one character splits into two or three components and several characters touch each other. Some methods are proposed to solve these problems. But, heuristically driven pitch information of characters is not competent enough for solving splitting and touching problems occurred in documents with various sizes and styles of characters. Although multistage graph search algorithm using dynamic programming can improve the segmentation results, but this method needs combinatorically increasing computing time. This chapter describes a character segmentation method using the matching rate between an input character and two finally selected candidate characters in the documents which consist of alphanumeric, symbols, Korean, and Chinese characters. The method can determine the exact cutting and merging point one by one character, consequently it needs small computing time. The experimental results have proven that the proposed method is efficient and accurate enough to enhance the performance of document recognition system.