Deep Learning Approaches on Super-Resolution Image and Video – A Deep Review

Abstract

In recent years, deep learning has become the subject of researchers. Typical image processing challenge is the super resolution on the image and on the film. Super-resolution is the job to create high resolution images from images of low resolution. Several Deep Learning approach have been acknowledged in this paper that is promoted for the improvement of the given input images. The major motivation of this learning for acquiring the information and observation of the various aspects of the deep learning approaches, which is broadly used in solving the complex problems and it overcomes the limitations of the traditional machine learning methods.  In this paper, before learning algorithms, the different characteristics of the SR approaches are defined. The models were explored in different super resolutions and contrasted with the PSNR (peak-signal noise ratio) and the SSIM measurements. Overwhelming on comparisons of the methods, Generative adversarial networks provides the finer solution in terms of the quality measurements. Therefore, this paper is furnished with the super-resolution image and video models of the image processing.