Image Registration for Super-Resolution using SIFT Key-points

Super-Resolution algorithms reconstruct a high resolution image from a set of low resolution images of a scene. An accurate image registration is a fundamental stage in all Super-Resolution methods. In this paper a new restriction criteria for matching SIFT key points, for registration under the condition of Super-Resolution problems is proposed. This is based on the assumption of normal distribution for location and orientation differences of correct matches under global translational motion model. The accuracy of the proposed method is compared with some famous registration methods. A quantitative comparison has been done with a collective of images with known motion vectors. Also a subjective comparison has been done with the results of various registration methods including the proposed method for Super-Resolution. The experimental results show the better performance of the proposed algorithm in compare with some other registration methods.