Ultra-low-bitrate image compression achieves 50% bitrate savings by treating decoding as a 'next-frame' video prediction task using diffusion priors.
March 17, 2026
Original Paper
Next-Frame Decoding for Ultra-Low-Bitrate Image Compression with Video Diffusion Priors
arXiv · 2603.15129
The Takeaway
It shifts the generative compression paradigm from static image priors to temporal video priors, significantly improving both visual realism and bitrate efficiency. This is a massive leap for high-fidelity communication in bandwidth-constrained environments.
From the abstract
We present a novel paradigm for ultra-low-bitrate image compression (ULB-IC) that exploits the ``temporal'' evolution in generative image compression. Specifically, we define an explicit intermediate state during decoding: a compact anchor frame, which preserves the scene geometry and semantic layout while discarding high-frequency details. We then reinterpret generative decoding as a virtual temporal transition from this anchor to the final reconstructedthis http URLmodel this progression, we l