Research map / Optimization / Nonconvex and smooth optimization
Plug-and-Play Methods Provably Converge with Properly Trained Denoisers
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train CNN denoisers with real spectral normalization rather than ordinary training (ablation) | SimpleCNN and DnCNN denoisers in PnP · PnP convergence experiments | denoiser residual Lipschitz bound ε | Helped |
| Use realSN-DnCNN instead of BM3D | PnP-ADMM · Poisson denoising test set | average PSNR (dB): 23.4617 -> 23.5873 | Helped |
| Use realSN-SimpleCNN instead of BM3D | PnP-FBS · Poisson denoising test set | average PSNR (dB): 18.5835 -> 22.7280 | Helped |
| Use PnP-ADMM rather than PnP-FBS with BM3D | BM3D plug-in denoiser · Poisson denoising test set | average PSNR (dB): 18.5835 -> 23.4617 | Helped |
| Use PnP-ADMM rather than PnP-FBS with RealSN-DnCNN | RealSN-DnCNN plug-in denoiser · Single photon imaging, best overall | average PSNR (dB): 28.4065 -> 31.0431 | Helped |
| Use RealSN-DnCNN instead of BM3D | PnP-ADMM · Single photon imaging, best overall | average PSNR (dB): 30.0474 -> 31.0431 | Helped |
| Use RealSN-DnCNN instead of DnCNN without realSN (ablation) | PnP-FBS · CS-MRI, 30% sampling | PSNR (dB): 19.59 -> 19.82; 16.49 -> 16.60; 18.92 -> 18.96; 15.99 -> 16.09; 14.76 -> 14.82; 14.09 -> 14.25 | Helped |
| Use RealSN-SimpleCNN instead of SimpleCNN without realSN (ablation) | PnP-FBS · CS-MRI, 30% sampling | PSNR (dB): 15.58 -> 17.65; 12.19 -> 14.98; 15.06 -> 16.52; 12.02 -> 14.26; 12.78 -> 13.02; 10.80 -> 11.49 | Helped |
| Use RealSN-DnCNN instead of DnCNN without realSN (ablation) | PnP-ADMM · CS-MRI, 30% sampling | PSNR (dB): 19.86 -> 19.91; 17.05 -> 17.09; 19.00 -> 19.08; 16.64 -> 16.68; 14.86 -> 15.11; 14.14 -> 14.16 | Mixed |
| Use RealSN-SimpleCNN instead of SimpleCNN without realSN (ablation) | PnP-ADMM · CS-MRI, 30% sampling | PSNR (dB): 16.68 -> 17.77; 12.56 -> 14.89; 16.83 -> 17.00; 13.47 -> 14.47; 13.03 -> 12.73; 11.17 -> 11.88 | Mixed |
| Replace PnP-FBS with PnP-ADMM using the same denoiser | PnP frameworks · CS-MRI, 30% sampling | PSNR (dB): BM3D: 19.09 -> 19.61, 16.36 -> 17.23, 18.10 -> 18.94, 15.67 -> 16.70, 14.37 -> 14.91, 12.99 -> 13.98; RealSN-DnCNN: 19.8 | Mixed |
| Use RealSN-SimpleCNN instead of RealSN-DnCNN | PnP-FBS · Poisson denoising test set | average PSNR (dB): 22.2154 -> 22.7280 | Helped |
Closest papers by meaning
- Plug-and-Play image restoration with Stochastic deNOising REgularization (ICML 2024 · similarity 0.90)
- Prior Mismatch and Adaptation in PnP-ADMM with a Nonconvex Convergence Analysis (ICML 2024 · similarity 0.88)
- PnP-Flow: Plug-and-Play Image Restoration with Flow Matching (ICLR 2025 · similarity 0.87)
- Learning Cocoercive Conservative Denoisers via Helmholtz Decomposition for Poisson Imaging Inverse Problems (NeurIPS 2025 · similarity 0.86)
- Reconciling Stochastic and Deterministic Strategies for Zero-shot Image Restoration using Diffusion Model in Dual (CVPR 2025 · similarity 0.86)
- Provably Robust Score-Based Diffusion Posterior Sampling for Plug-and-Play Image Reconstruction (NeurIPS 2024 · similarity 0.85)
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.