New neural video codec DCVC-MB leverages state space models for improved compression

Researchers have introduced DCVC-Mamba (DCVC-MB), a novel neural video codec framework designed for B-frame coding. This approach integrates an Interval Bound Propagation (IBP) strategy for low-delay B-frame compression and utilizes a spatio-temporal fusion model based on state-space models for bidirectional temporal prediction. Additionally, it features an entropy-aware skipping mechanism to reduce entropy coding times by omitting certain latents. Experiments show DCVC-MB outperforms previous neural video codecs and traditional codecs, achieving significant BD-rate reductions. AI

IMPACT This research advances neural video compression techniques, potentially leading to more efficient video encoding and streaming solutions.

RANK_REASON The cluster contains a research paper detailing a new method for neural video compression. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• DCVC-Mamba
• DCVC-MB
• Interval Bound Propagation
• Mamba
• VTM-19.0-LDP
• VTM-19.0-RA(Inter-GoP=16)

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →