适用于汉字识别的特征描绘子改进研究  

Improved Descriptor for Chinese Character Recognition

作　　者：武桐[1] 齐开悦[1] 陈凯[1] 陈剑波[1] 

机构地区：[1]上海交通大学信息安全工程学院,上海200240

出　　处：《信息安全与通信保密》2010年第3期79-81,共3页Information Security and Communications Privacy

摘　　要：提出了一种适用于汉字识别的新的局部描绘子,在SIFT描绘子的基础上添加了灰度对比直方图的概念。该局部描绘子作为汉字的特征表述,测试系统思想是把汉字识别当作图像匹配问题。为了检测新描绘子的特性,在仿射、噪声、背景、复杂情况这4种情况下进行汉字识别测试,对比新描述子和SIFT描绘子的汉字识别正确率,实验结果证明该描绘子在汉字识别方面可以描述更多的局部信息,表现也比SIFT描绘子要好。The paper proposes a new local descriptor for Chinese character recognition,which adds contrast gray histogram to SIFT descriptor.The proposed descriptor could be used in feature description of Chinese characters.The method of test system in this paper treats the character recognition as the image matching problem.To detect the property of the new descriptor,Chinese character recognition is tested under the four situations of affine,noise,background and complicated circumstance,and the right rate of Chinese character recognition is compared between the proposed descriptor and SIFT.Experiments indicate that the proposed descriptor could describe more local information and perform better than SIFT in Chinese character recognition.

关 键 词：尺度不变特征变换 描绘子 灰度对比直方图 匹配 汉字识别 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]