Shape Preserving and Noise Removing Image Enhancer through Regularization

摘要

This project concerns the development of an image reconstruction algorithm which removes noise from a corrupted image and recovers patterns and textures. The textures this algorithm detects are edges and smooth, gray scale regions. In previous attempts of this kind, images were recovered with nice edges, but blockiness remained in the smooth regions. Our goal is to recover both types of image composition. We do this by a least squares minimization method and a regularization process. The regularization is the key element that allows us to differentiate the edge and the smooth regions. The computation generated by the algorithm tends to create a large system to solve. We will use an algebraic multigrid solver to minimize the time for solving these large systems.