An image registration algorithm based on SIFT and CCH

Proper feature descriptors are critical in the processing of image registration. In this paper, to overcome the high dimension of the local descriptor and the lack of gray-level information of SIFT, an improved registration algorithm is proposed to combine the advantages of SIFT and CCH. In the new algorithm, the descriptor construction method is improved, gray-level information is added and the size of descriptor is reduced. Experimental results and comparative studies show that the new algorithm is invariant to image rotation and scaling and robust to illumination change as well as viewpoint change. And the new algorithm has high matching accuracy with less computation time.