基于极小阈值和曲线拟合的垂直投影汉字切分  

Vertical Projection Characters Segmentation Based on Minimum Threshold and Curve-Fitting

作　　者：王江晴[1] 曹卫[1] 

机构地区：[1]中南民族大学计算机科学学院,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2011年第4期82-85,共4页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(60975021)

摘　　要：针对传统汉字切分方法中对粘连汉字和重叠汉字切分不准确的问题,提出了基于极小阈值和曲线拟合的垂直投影汉字切分算法.该算法在传统垂直投影法汉字切分的基础上结合极小阈值和曲线拟合进行切分,实验结果表明:相对于传统汉字切分方法,该算法对含有粘连和重叠现象的手写体汉字文本的切分是有效的,从而可使手写体汉字的切分准确率有较大的提高.Aiming to solve the disadvantages of traditional handwritten Chinese character segmentation method, especially for segmentation inaccurate problem of adhesion characters and overlap characters, this paper presents an algorithm of vertical projection characters segmentation based on minimum threshold and curve-fitting. This algorithm combines minimum threshold and curve-fitting with traditional segmentation method. Experimental results show that this algorithm is effective for adhesion characters and overlap characters, and it can improve segmentation accuracy of Chinese characters.

关 键 词：阈值 过切分 弱切分 切分块 曲线拟合 极小值点 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]