Image Super-Resolution Algorithms Based on Sparse Representation of Classified Image Patches

Abstract

At present,super-resolution algorithms based on sparse representation of image patches exploit single dictionary to represent the image patches,which can not reflect the differences of various image patches types.In this paper,a novel method based on sparse representation of classified image patches is proposed to overcome this disadvantage.In this method,image patches are firstly divided into smooth patches,different directional edge patches and irregular structure patches by local features.Then these classified patches are applied into training the corresponding high and low resolution dictionary pairs.During the reconstruction process,simple bicubic interpolation approach is used for smooth patches while edge and irregular structure patches are reconstructed from their corresponding dictionary pairs using orthogonal matching pursuit algorithm.Experiment results show that the proposed algorithm significantly improves the visual quality of the edges and has faster speed compared with other single dictionary methods.