Two-phase approach for recovering images corrupted by Gaussian-plus-impulse noise

Images are corrupted by more than one type of noise during its acquisition and transmission stages. A two-phase denoising approach is presented in the paper to restore images corrupted by both Gaussian noise and impulse noise. The first-Detection phase involves identifying outlier candidates affected by impulse noise. The second-Reconstruction phase involves using group sparse modeling to reconstruct the image from the random samples unaffected by impulse noise. Group sparse representation model groups together the similar patches in an image. A self-adaptive group dictionary learning method is adopted for representing the groups. The robust split-Bregman iterative algorithm is used to solve the optimization problem. Experimental results support the effectiveness of the proposed algorithm over current state-of-the-art schemes.