Research map / Video / Video segmentation and tracking
Omniscient Video Super-Resolution
Peng Yi, Zhongyuan Wang, Kui Jiang, Junjun Jiang, Tao Lu, Xin Tian, Jiayi Ma
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add hidden state from the past and LR frame from the future (hybrid vs iterative) (ablation) | IVSR · 20 val sequences | PSNR: 30.66 -> 31.10 | Improved |
| add hidden states from present and future (local omniscient) (ablation) | HVSR · 20 val sequences | PSNR: 31.10 -> 31.24 | Improved |
| make the framework global (bidirectional) after tuning (ablation) | LOVSR 4+2 · 20 val sequences | PSNR: 31.28 -> 31.52 | Improved |
| add precursor SR output via alpha=0.01 instead of Bicubic residual (ablation) | GOVSR · 20 val sequences | PSNR: 31.27 -> 31.36 | Improved |
| add precursor SR output via alpha=0.1 instead of Bicubic residual (ablation) | LOVSR · 20 val sequences | PSNR: 31.24 -> 31.26 | Improved |
| raise precursor loss weight alpha from 0.1 to 1 (ablation) | LOVSR · 20 val sequences | PSNR: 31.26 -> 31.17 | Hurt |
| allocate 4 blocks to precursor instead of 0 (ablation) | GOVSR · 20 val sequences | PSNR: 31.20 -> 31.52 | Improved |
| allocate 4 blocks to precursor instead of 6 (6+0) (ablation) | GOVSR · 20 val sequences | PSNR: 31.43 -> 31.52 | Improved |
| feed HVSR a fourth LR frame (ablation) | HVSR · 20 val sequences | PSNR: 31.10 -> 31.15 | Improved |
| use LOVSR instead of HVSR with the same 4 LR frames | HVSR4 · 20 val sequences | PSNR: 31.15 -> 31.28 | Improved |
| add LR frame I_{t-1} to the successor network (ablation) | LOVSR 4+2 · 20 val sequences | PSNR: 31.29 -> 31.28 | No clear effect |
| add hidden state H_t of the precursor to the successor (ablation) | GOVSR 4+2 Net_s · 20 val sequences | PSNR: 12.44 -> 31.52 | Improved |
Closest papers by meaning
- BasicVSR: The Search for Essential Components in Video Super-Resolution and Beyond (CVPR 2021 · similarity 0.88)
- Look Back and Forth: Video Super-Resolution With Explicit Temporal Difference Modeling (CVPR 2022 · similarity 0.88)
- PS-SR: Pseudo-Single-Step Video Super-Resolution via Speculative Diffusion (CVPR 2026 · similarity 0.87)
- Frame and Feature-Context Video Super-Resolution (AAAI 2019 · similarity 0.87)
- SkipSR: Faster Super-Resolution with Token Skipping (NeurIPS 2026 · similarity 0.85)
- Sequence Matters: Harnessing Video Models in 3D Super-Resolution (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
