An Adaptive Momentum Term Based Optimization Method for Image Restoration

Linear inverse problems (LIP) arise in a wide range of applications in signal/image processing and statistics. In this paper, we propose an adaptive momentum term based optimization method for LIP in image restoration. Firstly, gradient descent with momentum is employed to search for iteration solutions, which can enlarge iterative steps in favorable directions and avoid some specific local optimal points caused by noise. Then, an adaptive parameter selection strategy based Barzilai-Borwein is selected to accelerate the iteration and increase robustness of algorithm. Thirdly, backtracking line search is used to attain a sufficient decrease, which can ensure monotonicity of the objective function. Experiment results demonstrate that the proposed method can obtain high-quality image restoration, especially, the robustness and number of iterations outperform its competitors.