Registration Algorithm In The Image Super-resolution Reconstruction

Image super-resolution (SR) reconstruction is the process of producing a high-resolution (HR) image from a sequence of low-resolution (LR) images. These LR images should represent the same scence, but with a different relative motion. It overcomes the inherent resolution limitation by bringing together the additional information from each LR images. Generally, there are two serial steps in super-resolution imaging. The first is to estimate the motion parameters. The second is to apply the information obtained from the different registered images to the reconstruction of a sharp HR image. So registration plays a critical role in SR reconstruction. An error in registration translates almost directly into degradation of the resulting HR image. At present, the common and effective registration algorithm in image SR reconstruction is optical flow based method.This paper focuses on optical flow based image registration, and studies the approach to obtain a HR image from observed multiple warped, blurred, decimated and noisy LR images. The main contributions of the paper are as follows:1. Optical flow based image registration is studied. The classical algorithms, such as Lucas-Kanade and Horn-Schunck, are discussed, where the results of the optical flow field using different parameters are given. We propose a gobal constraint algorithm based on coupled partial differential equation, then we introduce local constraint so that the optical flow of the degraded image can also be estimated accurately. Finally, we give a adaptive selection for parameter p in new algorithm.2. SR reconstruction technique after registration is studied. A comprehensive review of the basic theories and techniques of SR is addressed. We propose a regularization method based on coupled partial differential equation to slove the ill-posed problem of SR reconstruction. The method can eliminate random noise in the smooth region, and also can preserve texture. Finally, we realize SR reconstruction on the basis of new registration algorithm and new reconstruction algorithm. The experiments show that the new algorithms used for the SR reconstruction can effectively improve the quality of the reconstruction image.