Solving RED via Weighted Proximal Methods

REgularization by Denoising (RED) is a recently introduced framework for solving inverse problems by incorporating state-of-the-art denoising algorithms as the priors. Actually, RED shows that solving inverse problems amounts to iterated denoising processes. However, the complexity of denoisers is generally high, which may lead to an overall slow algorithm for solving RED. In this paper, we apply a general framework named weighted proximal methods (WPMs) to address RED efficiently. Moreover, we also show two existing solvers (namely the fixed point and accelerated proximal gradient methods) for RED are two special variants of WPMs. Numerical experiments show that we can obtain a more efficient variant of WPMs for handling RED if an effective weighting is set.