Deep Video Coding for AOM/AV2

The primary aim of this project is to investigate and integrate novel deep learning architectures, video coding tools and perceptual loss functions in the development of the next generation opensource standard video codec; AOM/AV2.

These new tools are expected to significantly increase the performance of the compression algorithms in comparison to previous generation standards, and also current deep learning-based frameworks.

This research will also aim to reduce the computational complexity of the novel deep learning tools, in order to facilitate adoption of deep learningbased frameworks within the AOM/AV2 standard.

Successful completion of this PhD will yield valuable results that will both further the surrounding literature regarding deep learning-based video compression, and also provide industry partners with a direct route to the integration of deep learning algorithms within their video compression standards.