Multi-modal Image Registration Based on Modified-SURF and Consensus Inliers Recovery

Multi-modal image registration has been received significant research attention in past decades. In this paper, we proposed a solution for rigid multi-modal image registration, which focus on handling gradient reversal and region reversal problems happened in multimodal images. We also consider the common property of multi-modal images in geometric structure for feature matching. Besides the improvements in features extraction and matching step, we use a correspondences recovery step to obtain more matches, thus improving the robustness and accuracy of registration. Experiments show that the proposed method is effective.