Reproduce “Boosting neural video codecs by exploiting hierarchical redundancy” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Reza Pourreza, Hoang Le, Amir Said, Guillaume Sautiere, Auke Wiggers
Researchers at Qualcomm AI Research enhanced neural video codecs by integrating lightweight pixel-space predictors to leverage hierarchical redundancies in video data. Their approach achieved up to 38% bitrate savings in RGB and 34% in YUV420 compared to the SSF baseline, demonstrating improved rate-distortion performance and competitive YUV MS-SSIM.
Key insights
- Explicitly exploiting second-order redundancy through dedicated pixel-space predictors offers substantial bitrate savings in neural video codecs.
- Lightweight, modular predictors can significantly enhance codec performance without requiring a complete redesign, making them broadly applicable.
- Comprehensive evaluation in the YUV420 colorspace is essential for assessing the practical relevance and competitiveness of neural video codecs against traditional standards.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.