A novel adaptive weights proximity matrix for image registration based on R-SIFT

We present an efficient and robust algorithm for image registration, which can be used to cope with large geometric and intensity variations between pair of images. There are two main contributions in this paper. First, a new robust scale invariant feature transform descriptor (R-SIFT) is presented, which is invariant under affine transformation. The second contribution is the development of a novel proximity matrix with adaptive weights based on the R-SIFT descriptor, in which the elements of the matrix combine the geometric information of feature points with the gradient information of feature points’ neighborhood. Quantitative comparisons of our algorithm with the related methods show a significant improvement in the presence of large viewpoint changes, scale changes, and illumination contrast. Experimental results for remote sensing image registration show our method outperforms the related methods. Finally, the experimental results of the proposed method applied to the problem of change detection of earthquake induced barrier lake are presented, validating the proposed algorithm.