A Multi-stage Approach for Segmenting Handwritten Chinese Characters
This paper illustrates an approch of composite segmentation for handwritten Chinese characters.As far as the segmentation is concerned in this paper, the various writings of handwritten Chinese characters may be classified into isolated characters, overlapping characters and touching characters. Aiming at the three writing types, the approach includes the following stages: Firstly, elementary non-linear segmentation paths are initially located by adopting Viterbi algorithm and the isolated or overlapping characters are obtained from the disconnected blocks of the original characters string. Subsequently, the segmentation of touching characters is developed through the touching position based on the shortest path method. Finally, for merging the over-segmented characters in the former stages, A* search algorithm is employed to choose the global optimization of all the candidate segmention paths. From legal amount on bank checks the experimental results on 400 Chinese character strings can achieve a reasonable good accuracy, which show the proposed approach is very feasible and effective for off-line handwritten Chinese characters segmentation.
Handwritten Chinese character segmentation Viterbi algorithm Shortest path A* algorithm Dynamic programming
Rui Ma Yongquan Xia Jingyu Yang
the University of Nanjing Science and Technology, Department of Computer Science, Nanjing, Jiangsu, the Department of Computer Science, University of Nanjing Science and Technology, Nanjing,Jiangsu, 2 the Department of Computer Science, University of Nanjing Science and Technology, Nanjing, Jiangsu,
国际会议
武汉
英文
2007-09-21（万方平台首次上网日期，不代表论文的发表时间）