YODA: Yet Another One-step Diffusion-based Video Compressor
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866918271016501248 |
|---|---|
| author | Li, Xingchen Zhang, Junzhe Shi, Junqi Lu, Ming Ma, Zhan |
| author_facet | Li, Xingchen Zhang, Junzhe Shi, Junqi Lu, Ming Ma, Zhan |
| contents | While one-step diffusion models have recently excelled in perceptual image compression, their application to video remains limited. Prior efforts typically rely on pretrained 2D autoencoders that generate per-frame latent representations independently, thereby neglecting temporal dependencies. We present YODA--Yet Another One-step Diffusion-based Video Compressor--which embeds multiscale features from temporal references for both latent generation and latent coding to better exploit spatial-temporal correlations for more compact representation, and employs a linear Diffusion Transformer (DiT) for efficient one-step denoising. YODA achieves state-of-the-art perceptual performance, consistently outperforming traditional and deep-learning baselines on LPIPS, DISTS, FID, and KID. Source code will be publicly available at https://github.com/NJUVISION/YODA. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_01141 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | YODA: Yet Another One-step Diffusion-based Video Compressor Li, Xingchen Zhang, Junzhe Shi, Junqi Lu, Ming Ma, Zhan Image and Video Processing Computer Vision and Pattern Recognition While one-step diffusion models have recently excelled in perceptual image compression, their application to video remains limited. Prior efforts typically rely on pretrained 2D autoencoders that generate per-frame latent representations independently, thereby neglecting temporal dependencies. We present YODA--Yet Another One-step Diffusion-based Video Compressor--which embeds multiscale features from temporal references for both latent generation and latent coding to better exploit spatial-temporal correlations for more compact representation, and employs a linear Diffusion Transformer (DiT) for efficient one-step denoising. YODA achieves state-of-the-art perceptual performance, consistently outperforming traditional and deep-learning baselines on LPIPS, DISTS, FID, and KID. Source code will be publicly available at https://github.com/NJUVISION/YODA. |
| title | YODA: Yet Another One-step Diffusion-based Video Compressor |
| topic | Image and Video Processing Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.01141 |