Matching feature points and discarding false matching pairs

Abstract

The approach is to integrate some traditional matching methods of feature point, and proposes a matching method of the feature points based on the fuzzy similarity measure of the feature points. For obtaining the complete matching between two sets of the feature points that hold the fuzzy similarity measure, we employ Hungarian method. For discarding the false matches from the matching result using Hungarian method, we employ the constraint of the epipolar geometry. Finally some experimental results are reported, which show the good performance of our scheme.