摘要
特征提取是文字识别中很重要的环节,传统用于特征提取的方法有模版法、变化特征法、投影直方图法和几何矩特征法等。文章简要介绍和分析了这些传统的特征提取方法及其优缺点,同时,指出了由于藏文字符的特殊性,决定了传统用于特征提取的几种方法在多字体藏文字符特征提取中效果不好的现状,并提出了一种"外围轮廓笔划特征提取法",该方法用于提出多字体的一个外围轮廓笔划的共同特征,效果较好。
Extracting the character of letter is the most important part. There were several extracting the character of letter, such as project, matrix character. The paper introduced and analyzed the an vantages and disadvantages of the traditional character of extracting as well the special Tibetan symbol influenced the effect of extracting with traditional character. Therefore, contour with enclosed draw character of extracting method was mentioned.
基金
教育部2006年度高等学校重大工程培育基金项目"印刷体藏文文字识别技术研究及其实现"阶段性成果。基金号:706059
关键词
文字识别
特征提取
多种字体
外围轮廓笔划特征
identified language
extracting character
multishape letters
contour with enclosed draw character