An iterative approach to super resolution

Abstract

In this paper, we present a new system based on a pyramidal scheme intended to enhance the resolution of images. This process involves two stages: the training stage, in which a high-resolution image is interpreted through multiple resolutions, which are presented as "training data" in the format of a Gaussian Quadtree Quadtree. The corresponding tree will be saved in Data Patch Files, which are intermediate files, to help the system work faster. In the second step called application stage, a pyramidal quadtree will be generated from the input low-resolution image, in order to create a higher-resolution image. By rearranging a tree at locations and resolutions where the input image quadtree has similar color characteristics with learned data, the corresponding high-resolution data is constructed.