SkipSR: Faster Super Resolution with Token Skipping
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908585609396224 |
|---|---|
| author | Choudhury, Rohan Lin, Shanchuan Wang, Jianyi Chen, Hao Zhao, Qi Cheng, Feng Jiang, Lu Kitani, Kris Jeni, Laszlo A. |
| author_facet | Choudhury, Rohan Lin, Shanchuan Wang, Jianyi Chen, Hao Zhao, Qi Cheng, Feng Jiang, Lu Kitani, Kris Jeni, Laszlo A. |
| contents | Diffusion-based super-resolution (SR) is a key component in video generation and video restoration, but is slow and expensive, limiting scalability to higher resolutions and longer videos. Our key insight is that many regions in video are inherently low-detail and gain little from refinement, yet current methods process all pixels uniformly. To take advantage of this, we propose SkipSR, a simple framework for accelerating video SR by identifying low-detail regions directly from low-resolution input, then skipping computation on them entirely, only super-resolving the areas that require refinement. This simple yet effective strategy preserves perceptual quality in both standard and one-step diffusion SR models while significantly reducing computation. In standard SR benchmarks, our method achieves up to 60% faster end-to-end latency than prior models on 720p videos with no perceptible loss in quality. Video demos are available at https://rccchoudhury.github.io/skipsr/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_08799 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SkipSR: Faster Super Resolution with Token Skipping Choudhury, Rohan Lin, Shanchuan Wang, Jianyi Chen, Hao Zhao, Qi Cheng, Feng Jiang, Lu Kitani, Kris Jeni, Laszlo A. Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Diffusion-based super-resolution (SR) is a key component in video generation and video restoration, but is slow and expensive, limiting scalability to higher resolutions and longer videos. Our key insight is that many regions in video are inherently low-detail and gain little from refinement, yet current methods process all pixels uniformly. To take advantage of this, we propose SkipSR, a simple framework for accelerating video SR by identifying low-detail regions directly from low-resolution input, then skipping computation on them entirely, only super-resolving the areas that require refinement. This simple yet effective strategy preserves perceptual quality in both standard and one-step diffusion SR models while significantly reducing computation. In standard SR benchmarks, our method achieves up to 60% faster end-to-end latency than prior models on 720p videos with no perceptible loss in quality. Video demos are available at https://rccchoudhury.github.io/skipsr/ |
| title | SkipSR: Faster Super Resolution with Token Skipping |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2510.08799 |