脱机手写汉字识别中笔段提取算法研究

基于目前细化和特征点提取的实现方法,提出了改进的分组细化方法和远端拐点法。改进的分组细化法能够在细化过程中,根据分组数标记字符图像中分叉点的类型,为后续的拐点提取做准备。给出了一种快速提取汉字拐点的方法远端拐点法。实验结果表明,该方法能够较好地提取笔段,特征点提取的正确率达到98.6％。

Sub-stroke extraction research on the off-line hand-written recognition of Chinese characters

Based on the present realizing method of thinning and feature point extraction, an improved thinning algorithm based on groups and a far inflection-point method was proposed. The method of the improved thinning algorithm can mark the type of branch points in the character images, which is based on group numbers in the thinning process. This method can prepare for sequential sub-stroke extraction. The far inflection-point can provide the method, which could quickly extract a Chinese character inflection-point. Experimental results show that these proposed algorithms can fairly well extract sub-strokes, and the feature point extraction accuracy is 98.6％.