Super Resolution from a Single Image based on Total Variation Regularization
Hiroki Tsurusaki, Masashi Kameda, Prima Oky Dicky Ardiansyah · International Conference on Intelligent Systems · 2014
In this paper, we propose an approach to generate super-resolution from a single image using iterative total variation regularization. During iterations, multiple pre-defined regulation parameters generate images representing skeletons and textures resulted by subtracting the original from the corresponding skeletons. We create a hybrid texture from the resulted textures and eliminate its non-edge regions to generate a high-resolution magnified image. Our experimental results show that the proposed approach generates magnified images with fined textures over the existing methods.