Overview of Research in the field of Video Compression using Deep Neural Networks

Deep Neural Networks (DNN) have emerged in recent year as a best-of-breed alternative for performing various classification, prediction and identification tasks in images and other fields of study. In the last few years, various research groups are exploring the option to harness them to improve video coding with the primary purpose of reducing video compression rates while retaining same video quality. Evolving neural-networks based video coding research efforts are focused on two different directions: (1) improving existing video codecs by performing better predictions that are incorporated within the same codec framework, and (2) holistic methods of end-to-end image/video compression schemes. While some of the results are promising and the prospects are good, no breakthrough has been reported as of yet. This paper provides an overview of state-of-the-art research work, providing examples of few prominent publications that illustrate and further explain the different highlighted topics in the field of using DNNs for video compression. Our conclusion is that the benefits have not been fully explored yet and additional work is expected to accomplish the next generation, neural networks based codecs.