关键词: 图像恢复, 总广义变差, 阶梯效应, 原始-对偶算法

Abstract: In order to preserve the geometric structure of image effectively, a novel nonconvex second-order total generalized variation image restoration model is proposed. The proposed model introduces a nonconvex sparse regularization term, and it is similar to the [L0] norm, so the new model can protect the structure features better. In addition, for computing the new model effectively, the reweighted method and primal-dual method are presented. Experimental results show that the proposed model can get better results compared with the recent method.

Key words: image restoration, total generalized variation, staircase effect, primal-dual method