Image Deblurring using Split Bregman Iterative Algorithm

Abstract

This paper presents a new variational algorithm for image deblurring by characterizing the properties of image local smoothness and nonlocal self-similarity simultaneously. Specifically, the local smoothness is measured by a Total Variation method, enforcing the local smoothness of images, while the nonlocal self similarity is measured by transforming the 3D array generated by grouping similar image patches. A new Split Bregman-based algorithm is developed to efficiently solve the above optimization problem. Extensive experiments on image deblurring verify the effectiveness of the proposed algorithm.