Primal-Dual Method for Hybrid Regularizers-Based Image Restoration with Impulse Noise

Abstract

With the aim of improving the restoration accuracy, this article introduces a hybrid regularizers approach to recovering images corrupted by impulse noise. The proposed model closely incorporates the superiorities of two recently developed methods: the total generalized variation method and the wavelet frame-based method. Numerically, a highly efficient primal-dual algorithm is constructed to solve the minimization problem, which is derived from the canonical alternating minimization method and based on the Moreau decomposition. Eventually, in comparison with several well-developed numerical methods, simulation experiments are provided to demonstrate the effective performance and advantages of our proposed strategy for image reconstruction under impulse noise, in terms of both image quality assessment and visual improvement.