An Accelerated Two-Step Iteration Hybrid-norm Algorithm for Image Restoration

Yong Wang, Wenjuan Xu, Xiaoyu Yang, Qianqian Qiao, Zheng Jia, Quanxue Gao

Abstract: Linear inverse problem is an important solution frame to solve image restoration. This paper develops an accelerated two-step iteration hybrid-norm reconstruction algorithm, exhibiting much faster convergence rate and better image than iteration shrinkage/thresholding based L1 norm algorithm. In the proposed method, hybrid norm model is built for image restoration objective function. Two-step iteration accelerates objective minimization optimization. Two-step iteration hybrid-norm algorithm converges to a minimizer of hybrid-norm objective function, for a given range of values of its parameters. Numerical examples are presented to validate that the effectiveness of the proposed algorithm is experimentally confirmed on problems of restoration with missing samples.