Generalized Difference Coder: A Novel Conditional Autoencoder Structure for Video Compression

TL;DR

This paper introduces the generalized difference coder, a new conditional autoencoder structure for video compression that improves residual coding efficiency by 27.8% over standard autoencoders, with moderate complexity increase.

Contribution

It proposes a novel generalized difference coder based on information theory, enhancing residual compression in learning-based video codecs.

Findings

Achieves 27.8% average rate savings

Maintains less than 7% complexity overhead

Provides theoretical foundation for conditional coding

Abstract

Motion compensated inter prediction is a common component of all video coders. The concept was established in traditional hybrid coding and successfully transferred to learning-based video compression. To compress the residual signal after prediction, usually the difference of the two signals is compressed using a standard autoencoder. However, information theory tells us that a general conditional coder is more efficient. In this paper, we provide a solid foundation based on information theory and Shannon entropy to show the potentials but also the limits of conditional coding. Building on those results, we then propose the generalized difference coder, a special case of a conditional coder designed to avoid limiting bottlenecks. With this coder, we are able to achieve average rate savings of 27.8% compared to a standard autoencoder, by only adding a moderate complexity overhead of less than 7%.