Research map / Video / Video segmentation and tracking
Video Super-Resolution with Recurrent Structure-Detail Network
Takashi Isobe, Xu Jia, Shuhang Gu, Songjiang Li, Shengjin Wang, Qi Tian
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 |
|---|---|---|---|
| add hidden state adaptation (HSA) to SD-block model (ablation) | RSDN 7-128 · Vid4 (ablation) | PSNR (dB): 27.76 -> 27.79 | Improved |
| add HSA to one-stream model (ablation) | One-stream 7-256 · Vid4 (ablation) | PSNR (dB): 27.58 -> 27.65 | Improved |
| add HSA to two-stream residual model (ablation) | Two-stream 7-128 · Vid4 (ablation) | PSNR (dB): 27.64 -> 27.68 | Improved |
| replace two-stream residual blocks with SD blocks (information exchange) (ablation) | RSDN 7-128 without HSA · Vid4 (ablation) | PSNR (dB): 27.64 -> 27.76 | Improved |
| replace one-stream network with SD blocks (ablation) | RSDN 7-128 without HSA · Vid4 (ablation) | PSNR (dB): 27.58 -> 27.76 | Improved |
| split input into structure and detail in plain two-stream network (no exchange) (ablation) | Two-stream 7-128 without HSA · Vid4 (ablation) | PSNR (dB): 27.58 -> 27.64 | No clear effect |
| feed structure and detail components instead of image to SD block (ablation) | RSDN 7-128 with HSA · Vid4 (ablation) | PSNR (dB): 27.73 -> 27.79 | Improved |
| feed structure and detail components instead of image to two-stream block (ablation) | Two-stream 7-128 with HSA · Vid4 (ablation) | PSNR (dB): 27.70 -> 27.68 | No clear effect |
| add whole-frame loss term (gamma=1) (ablation) | RSDN 7-128 · Vid4 (ablation) | PSNR (dB): 27.73 -> 27.79 | Improved |
| balance structure and detail loss weights (1,1) instead of structure-heavy (1,0.5) (ablation) | RSDN 7-128 · Vid4 (ablation) | PSNR (dB): 27.56 -> 27.79 | Improved |
| balance structure and detail loss weights (1,1) instead of detail-heavy (0.5,1) (ablation) | RSDN 7-128 · Vid4 (ablation) | PSNR (dB): 27.77 -> 27.79 | Improved |
| increase SD blocks from 5 to 7 | RSDN · Vid4 (4x) | PSNR Y (dB): 27.61 -> 27.79 | Improved |
Closest papers by meaning
- Recurrent Back-Projection Network for Video Super-Resolution (CVPR 2019 · similarity 0.91)
- Frame-Recurrent Video Super-Resolution (CVPR 2018 · similarity 0.90)
- Frame and Feature-Context Video Super-Resolution (AAAI 2019 · similarity 0.89)
- Robust Video Super-Resolution With Learned Temporal Dynamics (ICCV 2017 · similarity 0.88)
- One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution (NeurIPS 2025 · similarity 0.86)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.86)
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.
