Research map / Video / Video segmentation and tracking
Robust Video Super-Resolution With Learned Temporal Dynamics
Ding Liu, Zhaowen Wang, Yuchen Fan, Xianming Liu, Zhangyang Wang, Shiyu Chang, Thomas Huang
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| use 3 frames instead of 1 frame (B2 vs B1) (ablation) | temporal adaptive network · Video SR 4x, six test videos (all frames, architecture ablation) | Average PSNR: 27.33 -> 27.51 | Improved |
| use 5 frames instead of 1 frame (B3 vs B1) (ablation) | temporal adaptive network · Video SR 4x, six test videos (all frames, architecture ablation) | Average PSNR: 27.33 -> 27.73 | Improved |
| add adaptive temporal aggregation T to B1,2 instead of straight averaging (ablation) | temporal adaptive network · Video SR 4x, six test videos (all frames, architecture ablation) | Average PSNR: 27.54 -> 27.62 | Improved |
| add adaptive temporal aggregation T to B1,2,3 instead of straight averaging (ablation) | temporal adaptive network · Video SR 4x, six test videos (all frames, architecture ablation) | Average PSNR: 27.71 -> 27.96 | Improved |
| use B1,2,3 + T instead of the single long-term branch B3 (ablation) | temporal adaptive network · Video SR 4x, six test videos (all frames, architecture ablation) | Average PSNR: 27.73 -> 27.96 | Improved |
| use B1,2,3 + T instead of B3 on penguin (ablation) | temporal adaptive network · penguin (all frames) | PSNR: 36.40 -> 36.65 | Improved |
| add optical flow alignment to raw LR frames (ablation) | SR with B3 · Video SR 4x, six test videos (alignment ablation, B3 only) | Average PSNR: 27.51 -> 27.64 | Improved |
| use rectified optical flow instead of conventional optical flow (ablation) | SR with B3 · Video SR 4x, six test videos (alignment ablation, B3 only) | Average PSNR: 27.64 -> 27.73 | Improved |
| use spatial alignment network instead of no alignment (ablation) | SR with B3 · Video SR 4x, six test videos (alignment ablation, B3 only) | Average PSNR: 27.51 -> 27.67 | Improved |
| use spatial alignment network instead of rectified optical flow (ablation) | SR with B3 · Video SR 4x, six test videos (alignment ablation, B3 only) | Average PSNR: 27.73 -> 27.67 | Hurt |
| replace baseline trained on LR frames with proposed video SR as preprocessing | YouTube Face classifier · YouTube Face identification | Top-1 accuracy: 0.442 -> 0.511 | Improved |
| replace ESPCN with proposed on HD videos | video SR · Ultra Video Group | Average PSNR: 39.52 -> 39.71 | Improved |
Closest papers by meaning
- Real-Time Video Super-Resolution With Spatio-Temporal Networks and Motion Compensation (CVPR 2017 · similarity 0.92)
- Look Back and Forth: Video Super-Resolution With Explicit Temporal Difference Modeling (CVPR 2022 · similarity 0.92)
- Time Without Time: Pseudo-Temporal Representation for Space-Time Super-Resolution (CVPR 2026 · similarity 0.91)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.89)
- BF-STVSR: B-Splines and Fourier---Best Friends for High Fidelity Spatial-Temporal Video Super-Resolution (CVPR 2025 · similarity 0.89)
- Learning Spatial Adaptation and Temporal Coherence in Diffusion Models for Video Super-Resolution (CVPR 2024 · similarity 0.89)
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.
