An improved SIFT matching algorithm based on geometric similarity
Jinke Li, Geng Wang · 2015
SIFT(Scale Invariant Feature Transform) algorithm is a widely used method in image matching for its good invariance to image translation, rotation and illumination changes. However, due to local similarity of some images, when matching the SIFT feature points, some false matches occur. In this paper, an improved SIFT matching algorithm based on geometric similarity is proposed. For correct matches, there exists a constant similarity ratio. So based on it, we can eliminate false matching points which do not meet this condition. Experiments showed that the improved matching algorithm could eliminate mismatches effectively and make the feature points matching more accurate.