科研之友
微信
新浪微博
Facebook
分享链接
摘要
Spectral matching (SM) is widely used in correspondence extraction or refinement. However, it dose not work well in the instance when the given potential correspondences are mostly correct (i.e. the corresponding assignment graph is a coarsely connected graph). In the paper, we present an absolute spectral matching (ASM) method to solve this problem. In ASM, (1) the absolute value of the principal eigenvector is adored; (2) a user-defined parameter is defined to control the refinement process. Experimental results show the effectiveness of ASM over RANSAC.
- 出版日期2014
- 单位厦门理工学院
