Research map / Image generation / Super-resolution
TUDSR: Twice Upsampling-Diffusion for Higher Super-Resolution
Zhiqiang Wu, Yitong Dong, Xian Wei
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 |
|---|---|---|---|
| split x4 SR into two x2 stages (M2N2) rather than single x4 stage (ablation) | TUDSR-S · RealSR x4 | MUSIQ: 69.2568 -> 70.2406 | Improved |
| split x4 SR into two x2 stages (ablation) | TUDSR-S · RealLQ250 x4 | MUSIQ: 70.4876 -> 71.9250 | Improved |
| split x4 SR into two x2 stages (ablation) | TUDSR-S · RealSR x4 | MANIQA: 0.6818 -> 0.6786 | Hurt |
| split x4 SR into two x2 stages (ablation) | TUDSR-S · RealLQ250 x4 | NIQE: 3.5207 -> 3.9440 | Hurt |
| use stage-2 LoRA only (N4) instead of two-stage (ablation) | TUDSR-S · RealSR x4 | MUSIQ: 28.5303 vs 70.2406 | Hurt |
| split x8 SR into x4 then x2 (M4N2) rather than single x8 stage (ablation) | TUDSR-S · RealLQ250 x8 | MUSIQ: 51.5389 -> 63.2129 | Improved |
| split x8 SR into x4 then x2 (ablation) | TUDSR-S · RealSR x8 | MUSIQ: 65.7316 -> 67.2225 | Improved |
| split x8 SR into x4 then x2 (ablation) | TUDSR-S · RealSR x8 | NIQE: 4.4895 -> 4.6839 | Hurt |
| split x8 SR into x4 then x2 (ablation) | TUDSR-S · DrealSR x8 | NIQE: 5.3613 -> 5.5417 | Hurt |
| use stage-2 LoRA only (N8) instead of two-stage (ablation) | TUDSR-S · RealLQ250 x8 | MUSIQ: 19.6954 vs 63.2129 | Hurt |
| compare TUDSR-S with OSEDiff on x8 SR | TUDSR-S vs OSEDiff · RealLQ250 x8 | MUSIQ: 54.8024 -> 63.2129 | Improved |
| compare TUDSR-S with strongest one-step baseline on x4 | TUDSR-S vs PiSA-SR · RealSR x4 | FID: 124.19 -> 111.42 | 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.
- SeeSR — SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution (CVPR 2024) · better on 2 of 2 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
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.89)
- Noise-Started One-Step Image Super-Resolution via LR-Conditioned SplitMeanFlow and GAN Refinement (NeurIPS 2026 · similarity 0.89)
- Adversarial Diffusion Compression for Real-World Image Super-Resolution (CVPR 2025 · similarity 0.89)
- TSD-SR: One-Step Diffusion with Target Score Distillation for Real-World Image Super-Resolution (CVPR 2025 · similarity 0.89)
- DiffuseHigh: Training-Free Progressive High-Resolution Image Synthesis Through Structure Guidance (AAAI 2025 · similarity 0.88)
- UltraPixel: Advancing Ultra High-Resolution Image Synthesis to New Peaks (NeurIPS 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.
