基于数学形态学的脱机手写体汉字识别方法  

Off-line recognition method for handwritten Chinese character based on mathematical morphology

作　　者：吴媛[1] 杨扬[1] 颉斌[1] 王宏[1] 

机构地区：[1]北京科技大学信息工程学院,北京100083

出　　处：《计算机应用》2006年第3期622-623,626,共3页journal of Computer Applications

摘　　要：笔画特征是一种有效的脱机手写体汉字的识别特征,但是笔画细化往往会造成字体的变形,提出一种无需细化预处理的笔画特征提取方法,通过数学形态学中的腐蚀、膨胀等运算,采用不同的、具有自适应性的结构元素对汉字图像进行笔画分解,并利用弹性网格提取其方向特征,最后使用广义K-L变换对特征向量的维数进行压缩,去除冗余信息。实验结果验证了本方法的有效性。The stroke feature is efficient for off-line handwritten Chinese character recognition. But the process for thinning strokes always makes the characters distorted. A stroke feature extraction approach that doesn＇t need the thinning preproeoss was proposed. By erosion and dilation operations of the mathematical morphology and using different adaptive structure elements, the characters were decomposed into strokes, Then the elastic meshes were applied to extract the directional feature. Finally, generalized K-L transform was used to decrease the dimensions of the feature vectors so as to get rid of redundancies. Experiment results show the effectiveness of the proposed approach.

关 键 词：数学形态学 脱机手写体汉字识别 特征提取 细化预处理 广义K-L变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]