Super-Resolution Image Reconstruction for Gaussian Plus Salt-and-Pepper Noise Removal

A variational approach to reconstruct superresolution image corrupted by Gaussian and salt-and-pepper noise is studied. Since the salt-and-pepper noise is the outliers in the image, it is reasonable to regularize the data-fitting term by L1-norm. Full variational approach and two-phase approach for the data-fitting term are considered. To preserve the edges in the restored image, total variation norm is used as the regularization term. Subgradient method is applied to solve the optimization problem. Four difference iterative algorithms are tested and compared.