A Majorization-Minimization approach to total variation reconstruction of super-resolved images

Super-resolution is the task of reconstructing high-resolution images from shifted, rotated, low-resolution degraded observations. It can be formulated as an inverse problem for which regularization is necessary. In this paper we adopt this formulation and use the Total-Variation criterion for regularization. Then, we employ the Majorization-Minimization (MM) methodology to reconstuct high-resolution images from low-resolution observations. Experimental results are shown, which demonstrate the advantages of the proposed algorithm compared to other methods.