Parallel domain decomposition based algorithm for large scale color image denoising

An effective method for recovering blocky and discontinuous images from noisy data is the Total variation (TV) method. In this paper, we introduce a parallel domain decomposition based Newton-Krylov-Schwarz (NKS) method to numerically solve the total variation minimization model for the color image restoration. We show numerically that the NKS method converges and scales well on parallel computers with over one thousand processors. Several numerical experiments and comparisons demonstrate that the proposed method is fast and robust, especially for large scale color images.