Total-variation-type Image Restoration Problem Based On The Image Features

Image processing technology is the e?ective analysis and processing for degraded images, With the rapid development of it, Image processing gradually become an important way that people understand the world. Image processing mainly includes the related?elds, such as image restoration, image segmentation, image enhancement, etc. Image restoration is based on the degraded image and some prior knowledge, and designs some e?ect mathematical models, then ?ts degraded image to reverse the information of original image as much as possible. We research on the image restoration in this dissertation.In chapter 1, we introduce the research background and present situation of the digital image processing and image restoration.In chapter 2, we brie?y review some preliminaries and symbolic notations.In chapter 3, we mainly discuss the image restoration method based on total-variation regularized strategy. The image pixel values usually represent physical energy of the image and are required to be non-negative in theory, we introduce the bilateral constrained TV-L2 model. For this model, we propose to use the alternating direction method of multipliers to solve it and the regularization parameter is adjusted adaptively by the discrepancy principle. The numerical experiments show that the algorithm is e?cient in restoring blurred images that are corrupted by noise.In chapter 4, to image deblurring problem, we introduce the bilateral constrained image deblurring model based on high-order total-variation regularization and propose the alternating Bregman method to solve it. The numerical experiments show that the model and algorithm can obtain e?cient processed result.In chapter 5, we brie?y summarize our work in this paper and the direction of further research.