作者: Kjersti Aas , Line Eikvil , Tove Andersen
DOI: 10.1007/3-540-60268-2_335
关键词:
摘要: The problems of character recognition are today mainly due to imperfect thresholding and segmentation. In this paper a new approach text is presented which attempts avoid these by working directly on grey level images treating an entire word at the time. features found from levels image, hidden Markov model defined for each character. During most probable combination models use dynamic programming.