一些手写体汉字识别系统的笔划提取大都采取基于像素跟踪的细化算法,这些方法具有很大的不稳定性。考虑到汉字的结构特点,我们提出了汉字的直线段描述算法,直线段有0, 45, 90 和 135四种方向。在算法中,充分采用了基于结构的差分方法和基于统计的概率估计方法。实验表明,该算法有很好的鲁棒性,这为汉字的识别创造了条件。
A subdivision algorithm based on pixel tracking is adopted in most of stroke extracting in some recognition system for handwritten Chinese character. These methods have great instability. In consideration of the structure characteristics of Chinese character, we propose a description algorithm for straightway. The straightway consists of 4 directions of 0, 45, 90 and 135. In this algorithm, the differential method based on structure and probability estimate method based on statistics is fully used. The expe...