Study of Image Super Resolution Algorithms

Abstract

Super image resolution (SR) is a group of image processing technologies used in computer vision to improuve the resolution of deteriored images. Deep learning approaches have made great progress in super image resolution in recent years. In this study, we to provide a regular overview of current improuvments in image super resolution techniques using deep learning methodologies. Namely, we will describe and implement three SR algorithms : SRCNN, SRGAN and CAR. The comparative srudy is done in terms of the computation of two critera peak signal to noise ratio (PSNR) and the structure similarity index (SSIM). The obtained results have demonstrate the efficiency of the three amgorithms especialy CAR algorithm