摘要
在研究SIFT特征配准算法基础上,针对SIFT特征描述符的区域性特征,采用马氏距离对SIFT算法误匹配点进行剔除,以减少错误匹配,进而提高图像配准的正确率,并应用于纹理图像的配准。
It is aimed at districted feature of SIFT feature descriptor by studying the SIFT feature registration algorithm. In order to reduce the error registration, the Mahalanobis dis- tance is used in SIFT algorithm. Part of the error registration points are eliminated by the improved algorithm. The precision of image registration is increased. And it is used in tex- ture image registration.
出处
《沈阳理工大学学报》
CAS
2012年第4期6-10,共5页
Journal of Shenyang Ligong University
基金
辽宁省高等学校科技计划项目(1810162)
关键词
图像配准
SIFT
马氏距离
image registration
Scale Invariant Feature Transform SIFT)
Mahalanobis distance