基于文字笔画方向直方图的文本图像文种识别-信息工程大学学报.PDF

Vo l. 12 No.2 第 12 卷第 2 期 信息工程大学学报 2011 年 4 月 Joumal of Information Engineering University Apr.2011 基于文字笔画方向直方图的文本图像文种识别 郭龙，平西建，周林 (信息工程大学信息工程学院，河南郑州 450002) 摘要:针对文本图像文种识别中特征提取速度和识别精度之间的矛盾，提出了一种基于文字笔 画方向直方图的文种识别方法，利用笔画方向直方图对不同文种文字的笔画方向分布差异进 行描述并提取特征，采用支持向量机对所提特征进行训练和分类，实现文字种类识别。在实验 中选用有质量退化的中、英、俄、日、韩、阿拉伯等 10 种不同语言文字文本图像。实验结果表 明，本方法运算速度快，有较高的识别准确率并对图像质量退化有较好鲁棒性。 关键词:文本图像;文种识别;笔画方向直方图;支持向量机 中固分类号:TP39 1. 1 文献标识码:A 文章编号:1671 -0673(2011)02 -0231 -07 Script Identification of Document Image 8ased on Stroke Direction Histogram GUO Long , PING Xi-jian , ZHOU Lin (Institute of Information Engineering , Information Engineering University ,Zhengzhou 450002 , China) Abstract: Considering the contradiction between the speed of feature extraction and accuracy of i- dentification results in script identification of document image , this paper proposes a new script iden- tification algorithm based on the difference of the stroke direction distribution , and defines the stroke direction histogram , which describes the distribution of the stroke direction effectively. The Support Vector Machine (SVM) is applied for training and classifying the features extracted based on the stroke direction histogram to identify scripts in different languages. Experiments have been per- formed upon degraded document images , which include ten kinds of languages (Chinese , Russian , English , Japanese , Korean , Arabic , etc). Experimental results confirm that the proposed algorithm can