Research map / Video / Video generation
Temporal-Consistent Video Restoration with Pre-trained Diffusion Models
Hengkang Wang, Yang Liu, Huidong Liu, Chien-Chih Wang, Yanhui Guo, Hongdong Li, Bryan Wang, Ju Sun
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 shared-seed noise prior with small per-frame residuals (ablation) | seed-space MAP baseline · DAVIS video super-resolution x4 | WE (x10^-2): 1.398 -> 1.057; PSNR 24.70 -> 26.09 | Improved |
| add progressive warping with optical flow refinement (ablation) | seed-space MAP baseline · DAVIS video super-resolution x4 | WE (x10^-2): 1.398 -> 0.943; PSNR 24.70 -> 27.14 | Improved |
| combine noise prior and progressive warping (ablation) | seed-space MAP baseline · DAVIS video super-resolution x4 | PSNR: 24.70 -> 27.95; WE 1.398 -> 0.725 | Improved |
| use 4 DDIM reverse steps instead of 2 (ablation) | Ours · DAVIS video super-resolution x4 | PSNR: 27.87 -> 27.95 | Improved |
| increase DDIM reverse steps from 4 to 10 (ablation) | Ours · DAVIS video super-resolution x4 | PSNR: 27.95 -> 27.70 | Hurt |
| restrict residual ball radius to 0.1 (ablation) | Ours · DAVIS video super-resolution x4 | PSNR: 27.95 -> 25.74 | Hurt |
| increase residual ball radius from 1.0 to 10.0 (ablation) | Ours · DAVIS video super-resolution x4 | PSNR: 27.95 -> 27.91 | No clear effect |
| raise residual learning rate to 0.01 (ablation) | Ours · DAVIS video super-resolution x4 | PSNR: 27.95 -> 27.19 | Hurt |
| lower residual learning rate to 0.0001 (ablation) | Ours · DAVIS video super-resolution x4 | PSNR: 27.95 -> 26.17 | Hurt |
| replace CG-based zero-shot methods with seed-space MAP framework on motion deblurring | video restoration · DAVIS temporal deconvolution with spatial deblurring | PSNR: 15.94 -> 26.97 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VISION-XL — VISION-XL: High Definition Video Inverse Problem Solver using Latent Image Diffusion Models (ICCV 2025) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution (NeurIPS 2025 · similarity 0.88)
- VideoGuide: Improving Video Diffusion Models without Training Through a Teacher's Guide (CVPR 2025 · similarity 0.88)
- SeedVR: Seeding Infinity in Diffusion Transformer Towards Generic Video Restoration (CVPR 2025 · similarity 0.87)
- Warped Diffusion: Solving Video Inverse Problems with Image Diffusion Models (NeurIPS 2024 · similarity 0.87)
- Learning Temporally Consistent Video Depth from Video Diffusion Priors (CVPR 2025 · similarity 0.87)
- TRIP: Temporal Residual Learning with Image Noise Prior for Image-to-Video Diffusion Models (CVPR 2024 · similarity 0.87)
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.
