A Method of Image Registration Based On Best Similarity of Local Geometric Figure

Image Registration is an important part of computer vision. We propose a method of image registration by obtaining best similarity of local geometric figure that utilizes opposite core difference (OCD) of corresponding local figure. This method gets initial matching after describing precisely SIFT points by constructing feature subspaces based on the detection of SIFT feature points. Then we describe the corresponding similarity by OCD of local figure constructed by SIFT points and choose the feature points that possess highest similarity measure as point set to compute projective transformation matrix opt T . Experiments have proved that the precision of the matrix opt T and the Image matching is at a high level.