Feature Point Matching with Matching Distribution

Abstract

Abstract. Most of the feature point matching techniques considers only the number of matches. The higher number of matches is, the better results are. However, reliability and quality of the matching is addressed in a few techniques. So, finding the good matches of the pairs of points from the two given point sets is one of the main issue of feature point matching. This paper presents new approach to obtain reliable and good matches. The high quality of matching can be achieved when the matches are spread all over the entire point set. Therefore, we proposed to use the distribution of the matches to verify the quality of the matching. The preliminary results show that our proposed algorithm significantly outperform SIFT even when the results of SIFT are enhanced using RANSAC.