End-to-End Rate-Distortion Optimization for Bi-Directional Learned Video Compression

TL;DR

This paper introduces an end-to-end optimized bi-directional hierarchical video codec that outperforms previous sequential learned codecs in rate-distortion performance.

Contribution

It is the first to optimize a hierarchical, bi-directional learned video codec end-to-end by accumulating cost over GOPs, improving compression efficiency.

Findings

Bi-directional GOP codec outperforms sequential learned codecs.

End-to-end optimization enhances rate-distortion performance.

Hierarchical coding benefits from bi-directional motion compensation.

Abstract

Conventional video compression methods employ a linear transform and block motion model, and the steps of motion estimation, mode and quantization parameter selection, and entropy coding are optimized individually due to combinatorial nature of the end-to-end optimization problem. Learned video compression allows end-to-end rate-distortion optimized training of all nonlinear modules, quantization parameter and entropy model simultaneously. While previous work on learned video compression considered training a sequential video codec based on end-to-end optimization of cost averaged over pairs of successive frames, it is well-known in conventional video compression that hierarchical, bi-directional coding outperforms sequential compression. In this paper, we propose for the first time end-to-end optimization of a hierarchical, bi-directional motion compensated learned codec by accumulating cost function over fixed-size groups of pictures (GOP). Experimental results show that the rate-distortion performance of our proposed learned bi-directional {\it GOP coder} outperforms the state-of-the-art end-to-end optimized learned sequential compression as expected.