Research map / Image generation / Fast diffusion sampling
Score-Based Diffusion Models as Principled Priors for Inverse Imaging
Berthy T. Feng, Jamie Smith, Michael Rubinstein, Huiwen Chang, Katherine L. Bouman, William T. Freeman
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 score-based prior instead of TV, PCA-Gaussian or RealNVP priors | denoising with DPI posterior · Denoising CelebA test image (sigma=0.2) | PSNR: 20.5 (PCA-G) -> 25.1 | Improved |
| use mismatched score-based prior (trained on other dataset) (ablation) | denoising · Denoising CIFAR-10 test image | PSNR: 21.0 -> 19.0 | Hurt |
| use RealNVP normalizing flow as prior | denoising · Denoising CelebA test image | SSIM: 0.76 (PCA-G) -> 0.31 | Hurt |
| use variational posterior sampling with score-based prior instead of DPS, Score-ALD, SDE+Proj | deblurring · Deblurring CelebA image | average PSNR of posterior samples: 20.37 -> 24.75 | Improved |
| use score-based prior posterior sampling instead of baselines under interferometry | black-hole imaging · Interferometric imaging, black hole | max std. dev. of posterior samples: 0.202 -> 0.036 | Improved |
| use ODE-based gradient of log-probability instead of raw score-model output as gradient (ablation) | score-based prior · 16x16 Gaussian ground truth | cosine similarity to true gradient | Improved |
| tune measurement weight in baseline posterior samplers (ablation) | Score-ALD, DPS, SDE+Proj · 2D toy posterior | KL divergence to true posterior | No clear effect |
Closest papers by meaning
- Provably Robust Score-Based Diffusion Posterior Sampling for Plug-and-Play Image Reconstruction (NeurIPS 2024 · similarity 0.93)
- Whitened Score Diffusion: A Structured Prior for Imaging Inverse Problems (NeurIPS 2025 · similarity 0.91)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.91)
- Exact Posterior Score Estimation for Solving Linear Inverse Problems (NeurIPS 2026 · similarity 0.91)
- Rethinking Diffusion Posterior Sampling: From Conditional Score Estimator to Maximizing a Posterior (ICLR 2025 · similarity 0.90)
- Provable Posterior Sampling with Denoising Oracles via Tilted Transport (NeurIPS 2024 · 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.
