붙은 글자들이 포함된 인쇄체 한.영 혼용 문서에서의 효과적인 문자 인식 알고리즘

An Efficient Character Recognition Algorithm in Printed Korean/English Documents Including Touching Characters

In this paper, we present a character recognition algorithm in printed korean and english documents including touching characters. We derived two rules to segment and recognize touching characters in the bilingual documents, one from the shape characteristics of korean and english characters of the writing blocks defined in this paper, and the other from the RF (reliability factor) values generated from the classifiers. Overall classification accuracy for the KITE paper of the proposed algorithm was about 96.8% for the english abstract, and about 97.8% for the bilingual parts. Also we confirmed the proposed algorithm significantly improves the accuracy of character segmentation of the actual mixed korean and english documents including touching characters.