Rapid multimodality registration based on MM-SURF

With a large number of registration algorithms proposed, image registration techniques have achieved rapid development. However, there still exist many deficiencies in multimodality registration where high speed and accuracy are difficult to simultaneously achieve for real-time processing. In order to solve these problems we propose a novel method named MM-SURF (Multimodal-SURF). Inheriting the advantages of the SURF, the method is able to generate a large number of robust keypoints. For each keypoint, the neighborhood gradient magnitude is utilized to compute its dominant orientation. Relying on the dominant orientation, a MM-SURF descriptor is constructed as the local features description of the keypoint. The geometric transformation matrix for multimodal image registration is obtained by matching the keypoints. The method makes full use of gray information of multimodal images and simultaneously inherits the good performance of the SURF. Experimental results indicate that the proposed method achieves higher accuracy and consumes less runtime than the other similar algorithms for multimodal image registrations, and also demonstrate its robustness and stability in the presence of image blurring, rotation, noise and luminance variations.