Research map / Video / Video segmentation and tracking
Motion-Guided Latent Diffusion for Temporally Consistent Real-world Video Super-resolution
Xi Yang, Chenhang He, Jianqi Ma, Lei Zhang
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 motion-guided diffusion sampling (MDS) (ablation) | MGLD-VSR baseline · VideoLQ | warping error (WE): 9.3069 -> 8.9848 | Improved |
| add temporal-aware sequence decoder (TSD) (ablation) | MGLD-VSR baseline · VideoLQ | warping error (WE): 9.3069 -> 8.3679 | Improved |
| combine MDS and TSD (ablation) | MGLD-VSR · VideoLQ | warping error (WE): 9.3069 -> 7.5478 | Improved |
| use RAFT instead of SpyNet flow in motion-guided sampling (ablation) | MGLD-VSR · VideoLQ | warping error (WE): 7.6734 -> 7.5478 | No clear effect |
| use large motion guidance strength eta (ablation) | MGLD-VSR · VideoLQ | visual quality | Hurt |
| add CFW module (encoder feature) to decoder (ablation) | temporal-aware decoder · VideoLQ | warping error (WE): 9.8670 -> 8.6846 | Improved |
| add temporal modeling (1D conv) to decoder (ablation) | VAE decoder · VideoLQ | warping error (WE): 9.8670 -> 8.9842 | Improved |
| add frame difference loss and structure weighted consistency loss (ablation) | decoder fine-tuning · VideoLQ | warping error (WE): 7.9480 -> 7.5478 | Improved |
| use MGLD-VSR instead of RealBasicVSR for perceptual quality | RealBasicVSR · REDS4 | LPIPS: 0.4190 -> 0.3776 | Improved |
| diffusion prior method versus classic VSR on warping error | MGLD-VSR · REDS4 / UDM10 / SPMCS / VideoLQ | warping error (WE): REDS4 6.50 (RVRT) vs 10.08 (Ours) | Hurt |
| use MGLD-VSR (MIMO, lighter CFW) instead of StableSR per frame | StableSR · inference speed | time: 3.194s -> 1.113s | Improved |
Closest papers by meaning
- Rethinking Diffusion Model-Based Video Super-Resolution: Leveraging Dense Guidance from Aligned Features (CVPR 2026 · similarity 0.93)
- One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution (NeurIPS 2025 · similarity 0.92)
- Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution (CVPR 2024 · similarity 0.91)
- Self-supervised ControlNet with Spatio-Temporal Mamba for Real-world Video Super-resolution (CVPR 2025 · similarity 0.91)
- DOVE: Efficient One-Step Diffusion Model for Real-World Video Super-Resolution (NeurIPS 2025 · similarity 0.91)
- SkipSR: Faster Super-Resolution with Token Skipping (NeurIPS 2026 · similarity 0.90)
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.
