计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (34): 78-80.

• 学术探讨 • 上一篇    下一篇

一种新的特征描述符提取和匹配方法

王诗俊,杨树元   

  1. 中国科学院 声学研究所,北京 100080
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-12-01 发布日期:2007-12-01
  • 通讯作者: 王诗俊

New local descriptor extracting and matching approach

WANG Shi-jun,YANG Shu-yuan   

  1. Institute of Acoustics,Chinese Academy of Sciences,Beijing 100080,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-12-01 Published:2007-12-01
  • Contact: WANG Shi-jun

摘要: 特征描述符提取和匹配是基于特征点图像匹配方法的一个重要问题。针对特征描述符提取问题,分析了目前已提出的描述符提取方法存在的问题和不足,并针对这些问题提出一种新的描述符提取方法和相应的匹配方法,经过实验和比较,提出的方法有很高的准确率和稳定性。同时,还改进了特征点的梯度方向估计方法,使方向估计具有更高的准确率。

关键词: 特征点, 特征描述符, 特征匹配, 方向估计

Abstract: Extracting descriptors of interest points is a fundamental task in feature based image matching.In this paper,points out the shortcomings of some existing feature descriptors,and proposes a new feature descriptor extracting and matching approach.Performance evaluations show that this new descriptor provides higher distinctiveness and robustness to image rotate,scale and gray change.Meanwhile,proposes a new approach to compute dominant directions,which perform better than existing approaches.

Key words: interest points, local descriptor, matching, dominant direction