Research map / Video / Video segmentation and tracking
AVSR-Diff: Scale-Agnostic Diffusion Priors for Temporally Consistent Arbitrary-Scale Video Super-Resolution
Geunhyuk Youk, Jeonghyeok Do, Dayeon Kim, Jihyong Oh, Munchurl Kim
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 TGFR flow alignment to StableVSR baseline (ablation) | StableVSR · REDS4 (4x) | tLPIPS (x1e3): 5.40 -> 5.12 | Improved |
| add DCN sub-pixel refinement to TGFR (ablation) | TGFR · REDS4 (4x) | tLPIPS (x1e3): 5.12 -> 4.95 | Improved |
| swap pretrained decoder for continuous decoder without TGFR (ablation) | StableVSR · REDS4 (4x) | tLPIPS (x1e3): 5.40 -> 5.45 | Hurt |
| swap in continuous decoder after TGFR alignment (no SAFR) (ablation) | TGFR model · REDS4 (4x) | LPIPS (x1e2): LPIPS 9.71 -> 10.06; tLPIPS 4.95 -> 4.43 | Mixed |
| add SAFR to continuous video decoder (ablation) | AVSR-Diff · REDS4 (4x) | LPIPS (x1e2): 10.06 -> 9.54 | Improved |
| add gated fusion instead of concatenation in TGFR (ablation) | TGFR · REDS4 (4x), 100-frame pass | PSNR: 19.53 -> 28.75 | Improved |
| add gate sparsity penalty (ablation) | TGFR · REDS4 (4x), 100-frame pass | PSNR: 20.06 -> 28.75 | Improved |
| clip scale condition to s=4 at 8x inference (ablation) | AVSR-Diff · REDS4 (8x) | LPIPS (x1e2): 36.02 -> 29.43 | Improved |
| add scale-aware modules (with clipping) at 8x (ablation) | AVSR-Diff · REDS4 (8x) | LPIPS (x1e2): 31.01 -> 29.43 | Improved |
| use arbitrary-scale AVSR-Diff vs fixed-scale StableVSR + bicubic at large scale | AVSR-Diff · REDS4 (8x) | LPIPS (x1e2): 34.96 -> 29.43 | Improved |
| compare AVSR-Diff with regression-based RVRT on fidelity at 4x | AVSR-Diff · REDS4 (4x) | PSNR: 32.70 -> 28.75 | Hurt |
| denoise in fixed LR latent space vs VEnhancer's target-resolution denoising, 2x to 8x | AVSR-Diff · scale-wise efficiency | peak memory (GB): VEnhancer 40.3 -> 72.4 vs AVSR-Diff 8.7 -> 8.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- BF-STVSR — BF-STVSR: B-Splines and Fourier---Best Friends for High Fidelity Spatial-Temporal Video Super-Resolution (CVPR 2025) · better on 1 of 1 shared result
- SAVSR — SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024) · better on 1 of 1 shared result
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
- Stream-DiffVSR: Low-Latency Streamable Video Super-Resolution via Auto-Regressive Diffusion (ECCV 2026 · similarity 0.92)
- DiffVSR: Revealing an Effective Recipe for Taming Robust Video Super-Resolution Against Complex Degradations (ICCV 2025 · similarity 0.92)
- InfVSR: Toward Consistency-Driven Streaming Generative Video Super-Resolution (ICML 2026 · similarity 0.91)
- PatchVSR: Breaking Video Diffusion Resolution Limits with Patch-wise Video Super-Resolution (CVPR 2025 · similarity 0.90)
- One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution (NeurIPS 2025 · similarity 0.90)
- Upscale-A-Video: Temporal-Consistent Diffusion Model for Real-World Video Super-Resolution (CVPR 2024 · similarity 0.88)
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.
