基于笔划包围盒的脱机手写体汉字分割算法

Algorithm for offline handwritten Chinese character segmentation based on stroke bounding box

汉字分割是脱机手写体汉字识别预处理的重要部分,改进了基于笔画包围盒的汉字分割算法。根据方向行程长度提取出汉字的笔画,对所得笔画建立包围盒。使用4 种基本合并操作得到汉字的粗略的分割。最后使用动态规划的方法实现汉字的精细分割。实验结果表明,该算法较好地解决了汉字分割中的笔画粘连和重叠问题。

Chinese character segmentation is an important partof the preprocessing of offline handwritten Chinese character. An modified algorithm for Chinese characterssegmentation based on stroke bounding box is implemented. The length ofdirectional runs-based strokes are extracted. And then the bounding box for each stroke is constructed. A coarse segmentation of the character is obtained after four basic merging operation are applied. Finally, a dynamic programming method is used to find the optimal segmentation boundaries. Experiments show that the method can avoid the segmentation errors caused by the touched or overlapped strokes.