Adaptive total variation image restoration coupling with a gradient fidelity term

Abstract

Traditional total variation model often leads to an undesirable staircase effect. In this paper, a new image restoration model based on adaptive total variation is proposed to solve the problem. It couples adaptive total variation model with a gradient fidelity term. The validity of the model and uniqueness of the solution are demonstrated theoretically. In the algorithm processing, a more reasonable Laplace operator solution is used to keep better rotational invariance. Experimental results illustrate the coupled model not only preserves the edge and some important details but also avoids the staircase effect in smooth regions effectively.