Acta Scientiarum Naturalium Universitatis Pekinensis

Previous Articles     Next Articles

A New Automatic Seal Image Retrieval Method Based on Curves Matching

SONG Yong1, LIU Hong2   

  1. 1Center for Information Science, Peking University, Beijing, 100871;2National Key Lab. on Machine Perception, Peking University, Beijing, 100871
  • Received:2003-01-10 Online:2004-01-20 Published:2004-01-20

一种基于曲线匹配的印鉴图像自动检索方法

宋勇1, 刘宏   

  1. 1北京大学信息科学中心,北京,100871;2北京大学视觉与听觉信息处理国家重点实验室,北京,100871

Abstract: Chinese characters have many smooth vertical and horizontal strokes. According to that, two kinds of retrieval features in Chinese characters images are proposed and named as outer edge contour feature and orientation projection density feature. Further, by formularizing the two features with functional curves, matching of Chinese characters is transformed into matching of functional curves. Experiments show that these two kinds of features are stable and the curves matching based seal image retrieval method is effective.

Key words: Montgomery, RSA, VLSI Implementation

摘要: 从汉字笔划横平竖直这一特点出发,提出了汉字图像中适合于印鉴图像检索的两种特征:汉字的外边缘轮廓和汉字的方向投影密度。进而用函数曲线对这两种特征进行量化描述,将对汉字的匹配归结为对函数曲线的匹配。最后通过实验验证了这两种特征的稳定性以及基于这两种特征的曲线匹配方法的有效性。

关键词: 印章, 图像检索, 汉字特征, 曲线匹配

CLC Number: