Research On Multi-frame Image Super-resolution Reconstruction Algorithm

Multi-frame image super-resolution reconstruction is to fuse information from low-resolution images which have small displacement between each other for the same scene which need reconstruction. It produces high-resolution image by removing blur, noise and other factor which make image degradation in the process of fusion. Super-resolution reconstruction cost low, and it do not need to change hardware, so it has broad application prospects.This paper describes the background of super-resolution reconstruction algorithm, mathematical model, commonly used algorithm of super-resolution reconstruction, ill-posed problem and other related problem. Then as traditional super-resolution reconstruction algorithm is vulnerable to the effects of motion estimation errors which lead to decline in the quality of reconstruction. So I use the theory of robust estimation for super-resolution reconstruction algorithm framework to improve the robustness of the algorithm and the quality of reconstruction, In addition my method has adaptive parameter that can automatically adjust the weight of the input data. So it can not only effectively reduce outlier influence on the reconstruction, but also preserve the information better which is useful for reconstruction result.Besides, I also study joint motion estimation and super-resolution reconstruction algorithm. In this paper, I use the relationship of different low-resolution image degradation matrixes to construct a new object function of super-resolution reconstruction. Then, the super-resolution reconstruction process is divided into two parts in this paper:First, the input low-resolution image sequences do motion estimation and data fusion at the same time. And secondly, produce high-resolution image by removing blur, noise and other factor which make image degradation from the result of first part. I optimize motion estimation parameters in the process of super-resolution reconstruction and optimized motion estimation parameters in turn will enhance the accuracy of super-resolution reconstruction. Finally, experimental results prove my algorithm is superior to the traditional super-resolution reconstruction algorithm.