摘要： 本文提出了一种用计算机从手写汉字中抽提出三角号码基本特征的方法,并对抽出的汉字三只角上的特征进行编码,以达到识别分类的目的。从信息论的观点进行研究已经知道,汉字图像的四周对于文字识别的信息量是相对集中的,如果能正确地抽取这些特征,识别就成为可能。具体方法如下:(1)将手写汉字正规化,除去噪音,抽出方向线段,提取特征点的座标及接续关系矩阵。(2)在汉字图像上找出右上、左下、右下三个角上的基本特征点,根据特征点矩阵对每一个字生成三个笔形图像。(3)将笔形图像与三角号码笔形辞典比较,识别抽出三角号码笔形,并加以编码。用计算机模拟实验表明,本方法是有效的。

Abstract: Extraction of Three Corner Codes (TCC) from Chinese characters with a computer is proposed in the paper. The computer, to which an ability of recognition was given, can extract the features of TCC in order to recognize and catalog. The method is as follows: (1) The feature points and relation matrix of the Chinese characters were extracted without thinning process. (2) Depending on the feature points, three pictures of strokes were drawn, which correspond to the strokes in three corners of one character respectively. And (3) The pictures match with pictures of strokes in TCC dictionary. The experimental results with a computer show the satisfactory validity of the proposed method.