Research map / Image generation / Super-resolution
Bridging the Distribution Gap to Harness Pretrained Diffusion Priors for Super-Resolution
JoonKyu Park, Kyoung Mu Lee
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 adaptive predicted timestep instead of a fixed timestep (ablation) | DM-SR · RealSet80 | MUSIQ: 69.990 -> 70.616 (best fixed 250) | Improved |
| use larger fixed timestep (500 vs 50) (ablation) | DM-SR · RealSet80 | MUSIQ: 68.982 -> 69.078 | Improved |
| add adversarial loss to L1+perceptual (ablation) | DM-SR · RealSet80 | MUSIQ: 64.153 -> 69.947 | Improved |
| add distribution matching loss alone (ablation) | DM-SR · RealSet80 | MUSIQ: 64.153 -> 69.089 | Improved |
| add noise loss alone (ablation) | DM-SR · RealSet80 | MUSIQ: 64.153 -> 69.189 | Improved |
| add Ldm and Leps on top of Ladv (ablation) | DM-SR · RealSet80 | MUSIQ: 69.947 -> 70.616 | Improved |
| add Ldm on top of Ladv (ablation) | DM-SR · RealSR | LPIPS: 0.328 -> 0.317 | Improved |
| use LPIPS instead of L1 pixel distance as timestep ground truth (ablation) | DM-SR timestep estimator · RealSet80 | MUSIQ: 70.328 -> 70.616 | Improved |
| use LPIPS instead of SSIM as timestep ground truth (ablation) | DM-SR timestep estimator · RealSet80 | MUSIQ: 70.352 -> 70.616 | No clear effect |
| supervise timestep by min reconstruction error (min*) (ablation) | DM-SR timestep estimator · RealSet80 | MUSIQ: 70.616 -> 65.630 | Hurt |
| increase sampling steps from 1 to 5 (ablation) | DM-SR · RealSet80 | LIQE: 4.652 -> 4.773; MUSIQ 70.616 -> 69.948 | Mixed |
| increase sampling steps from 1 to 20 (ablation) | DM-SR · RealSet80 | LIQE: 4.652 -> 4.152 | Hurt |
Closest papers by meaning
- Taming Diffusion Prior for Image Super-Resolution with Domain Shift SDEs (NeurIPS 2024 · similarity 0.93)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.92)
- Unleashing the Power of One-Step Diffusion based Image Super-Resolution via a Large-Scale Diffusion Discriminator (NeurIPS 2025 · similarity 0.92)
- When Latents Forget Pixels: Restoring Fidelity in Diffusion Transformer Super-Resolution (NeurIPS 2026 · similarity 0.91)
- DREAM: Diffusion Rectification and Estimation-Adaptive Models (CVPR 2024 · similarity 0.91)
- Solving Diffusion ODEs with Optimal Boundary Conditions for Better Image Super-Resolution (ICLR 2024 · similarity 0.91)
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.
