Stroke Extraction of Handwritten Chinese Character Based on Ambiguous Zone Information
Zhengyang Zhou, Enqi Zhan, Jianbin Zheng · 2017
Stroke extraction plays an important role in the analysis of handwritten Chinese character. Ambiguous zones like intersections and junctions of strokes always bring difficulties for the extraction. The skeleton obtained by thinning algorithm is easy to be distorted in these areas. To solve this problem, an effective method to extract strokes using ambiguous zone information is proposed in this paper. This method uses an 8-neighbour window to detect ambiguous zones on the skeleton first. The removal of ambiguous zones splits the skeleton into several stroke segments. Cosine similarity and smooth degree between local sub-segments are calculated to get the connection coefficient. A reasonable strategy is made to connect stroke segments which satisfy connection coefficient conditions by interpolation. Natural strokes with dynamic information are extracted through the reconstruction and distortions on the skeleton are also modified. The last step detects abrupt turning points to get the final set of strokes. In the experiment, the proposed method obtains obviously better effect when detecting ambiguous zones. The extracted strokes remain good shapes and the position relationship among strokes can be reflected correctly. Results show that the stroke extraction method based on ambiguous zone information achieves high accuracy and fast speed, which can be useful in the related researches on handwritten Chinese character.