A Image Registration Method Combined with Gradient Similar Measure

Combined with the gradient similar measure, this paper proposes a new algorithm to register the images based on registration operator. By using the bilinear interpolation to express the relationship between the image to be registered and the reference image, the method transfers the registration problem to parameter estimation. And on the basis of the revised gradient similar measure, we construct the multiple-objective optimal model to register images. The registration parameters not only infer to shift parameters, but also expand to rotation parameters and scale parameter, thus enlarging the algorithm application. The optimal problem is resolved by Powell method. Finally the simulation and real image tests show that the method improves the registration precision based on rough registration. And because of the analytical expression between the image to be registered and the reference image, the method makes the postprocess superior, such as multi-frame images restoration and super-resolution.