CANF-VC++: Enhancing Conditional Augmented Normalizing Flows for Video Compression with Advanced Techniques

TL;DR

This paper introduces CANF-VC++, an improved learned video compression codec that integrates advanced techniques to achieve significant rate savings and outperforms existing standards like VTM on multiple datasets.

Contribution

The paper presents CANF-VC++, a novel enhanced codec that systematically incorporates new architectural and training techniques to improve compression efficiency over prior learned codecs.

Findings

Achieves significant Bj{}ntegaard-Delta rate savings on multiple datasets.

Outperforms baseline CANF-VC and H.266 VTM in coding efficiency.

Demonstrates the effectiveness of integrating advanced techniques into learned video codecs.

Abstract

Video has become the predominant medium for information dissemination, driving the need for efficient video codecs. Recent advancements in learned video compression have shown promising results, surpassing traditional codecs in terms of coding efficiency. However, challenges remain in integrating fragmented techniques and incorporating new tools into existing codecs. In this paper, we comprehensively review the state-of-the-art CANF-VC codec and propose CANF-VC++, an enhanced version that addresses these challenges. We systematically explore architecture design, reference frame type, training procedure, and entropy coding efficiency, leading to substantial coding improvements. CANF-VC++ achieves significant Bj{\o}ntegaard-Delta rate savings on conventional datasets UVG, HEVC Class B and MCL-JCV, outperforming the baseline CANF-VC and even the H.266 reference software VTM. Our work demonstrates the potential of integrating advancements in video compression and serves as inspiration for future research in the field.