Application of SIFT feature extraction algorithm on the image registration

Presents a high-accuracy image registration technique using SIFT and BBF. Conventional techniques of SIFT-based image registration are not very effective when searching for homonymy point-pairs, and eliminating mismatching. This paper proposes a technique to improve registration performance by BBF and RANSAC algorithm, BBF algorithm searches optimal homonymy point-pairs, and RANSAC algorithm eliminates mismatching. Experiments with real-world scenes demonstrate that the proposed method can reach a better precision and robustness, which outperforms previously proposed schemes, and makes possible to estimate image displacement with 0.01 pixel accuracy.