A New Algorithm for Image Registration and Super-Resolution Reconstruction

The accurate knowledge of the sub-pixel registration parameters for each observed Low-Resolution (LR) image is the key for image Super-Resolution Reconstruction (SRR). Many registration algorithms fail to provide sufficient precision for SRR when the LR images are severely aliased. However, the method which combines the registration problem into SRR can obtain accurate estimation for registration parameters of each LR image in this situation. This paper analyzed the method and proposed a new approach to solve it. The proposed approach uses the principle similar to variable projection which results in a better-conditioned problem and avoids some shortcomings of cyclic coordinate descent optimization procedure. It can be efficiently implemented by using Lanczos method and Gauss quadrature theory. As a result, the proposed approach can deal with translation and rotation between the LR images. Experimental results demonstrate the effectiveness of our approach.