Stroke-based thinning algorithm for Chinese characters

This paper proposes a stroke-based thinning algorithm for handwritten Chinese characters. One stroke which crosses or connects other strokes is divided into segments based on the information about the direction runs in the stroke. Four directions are considered and are represented by four direction codes. One of the four-direction codes is assigned to each foreground pixel. Connected pixels with the same direction code form a block. We split blocks into stroke and fork segments according to the direction codes of the blocks and the adjacent relationships among the blocks. The skeletons of the stroke segments are extracted; no skeleton point is extracted in the fork segments. Skeleton segments that connect to one fork segment are joined to reconstruct complete stroke skeletons by replacing the fork segment with line segments. The four directions are only used to split strokes; the thinned strokes still have any possible direction. The proposed approach is free from Y-shape distortion, L-shape distortion, the hairy problem, the shortening problem, and the hole problem. This approach is also insensitive to character noise and orientation. Experimental results show that the proposed approach produces better skeletons than do other general-purposed thinning methods for handwritten Chinese characters.