A Hybrid Deep Learning Technique for Compressed Video Enhancement

Abstract

In recent years, the high-quality video is highly demanded. But at the same time, video quality is degraded due to factors such as compression, noise, blur and various environmental conditions, which give rise to the need for using the various video enhancement techniques These techniques aim to recover the quality of degraded videos and therefore enhancing the visual quality of the videos by addressing tasks such as denoising, super-resolution, coloration , stabilization de-flickering, and reducing compression artifacts using traditional methods as well as modern methods ,which will be discussed in the paper later. Despite using notable advancements, several issues remain unresolved, and overcoming them is vital to achieving robust, efficient, and widely applicable solutions. Traditional video enhancement methods suffer from many limitations like not balancing computational efficiency and high-quality output, which give rise to degraded performance in real-time scenarios. Over the past few years, deep learning methods have brought substantial advancements in the enhancement of compressed video, prominently surpassing the traditional techniques in terms of enhancing the visual quality by reducing the compression effects manifolds. In this paper, an extensive overview of the video enhancement techniques is presented and a Hybrid Deep Learning Enhancement Technique for Compressed Video is proposed.