Video Quality Enhancement using Generative Adversarial Networks-based Super-Resolution and Noise Removal
Mobeen Ahmad, Muhammad Abdullah, Dongil Han · 2021 36th International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC) · 2021
Video quality enhancement is a challenging task as it not only involves super-resolution but also there is underlying noise in most of the real-world videos recorded almost a decade ago. Existing literature focuses on image super-resolution-based methods which fail to deliver satisfactory results in real-world scenario due to lack of high-resolution and low-resolution pairs. We propose a method based on image translation methodology coupled with super-resolution architecture. This does not require high-resolution, low-resolution pair, and learns the underlying noise automatically, Furthermore, it can learn the style of a High-Definition video and apply it on a low-resolution video. We present qualitative results that show excellent performance on unseen dataset.