A Review of Super Resolution Based on Deep Learning

Super-resolution (SR) is the process of restoring a limited number of low-resolution (LR) images to high-resolution (HR) images. In recent years, with the vigorous development of deep learning in computer vision, its applications in image super resolution have also made significant progress. In this paper we aim to integrate and analyze the existing deep-learning based image super-resolution models, and show several models with the best performance. We divide the models into five main categories based on where the sampling is located in the different models. Then we analyze and compare the different networks architectures. We also list some tips which are effective in super-resolution networks. At the end of this paper, we analyze the existing problems of super-resolution based on deep learning and make an outlook on the future of super-resolution development.