Research map / Image generation / Image restoration
Frequency-Guided Posterior Sampling for Diffusion-Based Image Restoration
Darshan Thaker, Abhishek Goyal, Rene Vidal
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 |
|---|---|---|---|
| use time-varying low-pass filtered measurement (FGPS) instead of DPS | DPS · FFHQ 256 high-pass deconv | FID: 121.57 -> 10.02 | Improved |
| use FGPS instead of DPS | DPS · ImageNet 256 high-pass deconv | FID: 111.73 -> 24.44 | Improved |
| use FGPS instead of DPS | DPS · FFHQ 256 motion deblur | FID: 27.15 -> 20.48 | Improved |
| use FGPS instead of DPS | DPS · ImageNet 256 Gaussian deblur | FID: 98.02 -> 56.46 | Improved |
| use FGPS instead of DPS | DPS · FFHQ 256 Gaussian deblur | FID: 21.69 -> 21.89 | No clear effect |
| use FGPS instead of DPS | DPS · FFHQ 256 Gaussian deblur | PSNR: 26.01 -> 25.44 | Hurt |
| use FGPS instead of DSG | DSG · FFHQ 256 dehazing | FID: 10.55 -> 9.37 | Improved |
| use FGPS instead of DPS | DPS · FFHQ 256 dehazing | LPIPS: 0.137 -> 0.066 | Improved |
| use FGPS instead of supervised AOD-Net | AOD-Net · FFHQ 256 dehazing | PSNR: 17.42 -> 26.88 | Improved |
| use time-dependent frequency curriculum instead of fixed low-pass filter (ablation) | FGPS · FFHQ high-pass deconv | reconstruction quality | Improved |
| raise the cutoff for k_T (final filter) (ablation) | FGPS · FFHQ | reconstruction quality | No clear effect |
| use exponential schedule for faces and uniform for ImageNet (ablation) | FGPS · FFHQ and ImageNet | reconstruction quality | 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.
- DoubleDIP — "Double-DIP": Unsupervised Image Decomposition via Coupled Deep-Image-Priors (CVPR 2019) · 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
- Improving Diffusion Posterior Samplers with Lagged Temporal Corrections for Image Restoration (NeurIPS 2026 · similarity 0.91)
- Analyzing and Guiding Zero-Shot Posterior Sampling in Diffusion Models (NeurIPS 2026 · similarity 0.90)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.90)
- DGSolver: Diffusion Generalist Solver with Universal Posterior Sampling for Image Restoration (NeurIPS 2025 · similarity 0.90)
- Improving Diffusion-Based Image Restoration with Error Contraction and Error Correction (AAAI 2024 · similarity 0.89)
- Ambient Diffusion Posterior Sampling: Solving Inverse Problems with Diffusion Models Trained on Corrupted Data (ICLR 2025 · similarity 0.89)
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.
