복잡한 컬러 문서에서의 문자추출 및 문자인식 = Text Location And Character Recognition on Complex color Documents

다국어 초록 (Multilingual Abstract)

This paper proposes character recognition system for complex color documents.
The proposed system consists of two components : text extraction and text recognition. It extracts text regions using a neural network that trains a set of texture discrimination masks for the given texture clases : text region and non-text region. And it extracts character-segments from the text regions using location and color information. Character recognition is performed by the comparison of the extracted character-segment and the templates that represent shapes of characters. To assess the validity, the experiments are performed with a variety of documents.