Video Compression through Image Interpolation

TL;DR

This paper introduces a novel end-to-end deep learning video codec that leverages image interpolation techniques, outperforming traditional codecs like H.261 and MPEG-4, and matching H.264 performance.

Contribution

It proposes a new deep learning-based video compression method based on image interpolation, replacing traditional hand-designed algorithms.

Findings

Outperforms H.261 and MPEG-4 codecs

Performs on par with H.264

Utilizes recent advances in deep image interpolation

Abstract

An ever increasing amount of our digital communication, media consumption, and content creation revolves around videos. We share, watch, and archive many aspects of our lives through them, all of which are powered by strong video compression. Traditional video compression is laboriously hand designed and hand optimized. This paper presents an alternative in an end-to-end deep learning codec. Our codec builds on one simple idea: Video compression is repeated image interpolation. It thus benefits from recent advances in deep image interpolation and generation. Our deep video codec outperforms today's prevailing codecs, such as H.261, MPEG-4 Part 2, and performs on par with H.264.