Total variation image restoration model by adaptively selecting parameter

Abstract

The choice of the suitable regularization parameter is very important to improve the quality and speed of the image inverse problem. To this end, this paper proposes a new scheme to choose the regularization parameter for the image restoration based on the total variation regularization model. To be more specific, since the proposed model is a nonconvex and nonsmooth optimization problem, we employ the alternating direction method of multipliers (ADMM) to split the original model into several easily solvable subproblems and then the regularization parameter can be efficiently computed by using the meantime Morozov deviation principle. Numerical results show that the proposed model and algorithm are efficient in restoring blurred images that are corrupted by noise.