基于数学形态学的联机手写字符识别去噪方法  

De-noising Approach for Online Handwriting Character Recognition Based on Mathematical Morphology

作　　者：孙嫣[1,2] 刘瀚猛[1] 芮建武[1] 吴健[1] 

机构地区：[1]中国科学院软件研究所基础软件国家工程研究中心,北京100190 [2]中国科学院研究生院,北京100049

出　　处：《计算机科学》2009年第10期237-239,295,共4页Computer Science

基　　金：信息产业部电子信息产业发展基金项目"跨平台的藏文办公套件研发"资助

摘　　要：手写输入时由于笔尖抖动等原因产生了大量噪声,有效地去除噪声是手写识别的前提和关键。根据联机手写识别中手写体字符形态的特性,分析了手写时由于各种原因而产生的噪声,运用数学形态学中膨胀、腐蚀、细化等基本运算,提出了一种将数学形态学应用于联机手写识别预处理的方法,该方法可以有效地消除大量的冗余信息。测试结果表明,提出的方法可行,具有很好的鲁棒性,可以配合其他方案应用于各种联机手写字符识别中。There are lots of various noises when users are writing characters on the tablet. It＇s significant to eliminate these noises in order to make these characters be recognized accurately. According to the features of online handwriting characters, we analyzed all sorts of noises generated during writing on the tablet. By applying erosion, dilation, thinning operations of the mathematical morphology into pre-processing of online handwriting recognition, a de-noising approach was proposed in this article. Using appropriate structure elements, we can eliminate large amounts of noises in the glyphs of characters. The experiment shows that this solution is valid and robust for online handwriting character recognition.

关 键 词：数学形态学 联机手写识别 噪声消除 细化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]