Research map / Image generation / Image restoration
Simultaneous Fidelity and Regularization Learning for Image Restoration
Dongwei Ren, Wangmeng Zuo, David D. Zhang, Lei Zhang, Ming–Hsuan Yang
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 |
|---|---|---|---|
| train with negative SSIM loss instead of MSE loss (ablation) | SFARL · Rain1400 | SSIM: 0.9153 -> 0.9188 | Improved |
| train with negative SSIM loss instead of MSE loss (ablation) | SFARL · Rain1400 | PSNR: 31.48 -> 31.37 | Hurt |
| use task-specific regularization filters instead of ones transferred from deconvolution (ablation) | SFARL · Rain12 deraining | PSNR: 29.66 -> 35.97 | Improved |
| use task-specific regularization filters instead of ones transferred from denoising (ablation) | SFARL · Rain12 deraining | PSNR: 32.02 -> 35.97 | Improved |
| use task-specific fidelity filters instead of ones transferred from deconvolution (ablation) | SFARL · Rain12 deraining | SSIM: 0.6769 -> 0.9581 | Improved |
| use task-specific fidelity filters instead of ones transferred from denoising (ablation) | SFARL · Rain12 deraining | SSIM: 0.7508 -> 0.9581 | Improved |
| replace IRCNN with SFARL (learned fidelity term) | deconvolution with inaccurate kernels · Levin, Xu and Jia kernels | SSIM: 0.9023 -> 0.9164 | Improved |
| replace IRCNN with SFARL (learned fidelity term) | deconvolution with inaccurate kernels · Levin, Cho and Lee kernels | SSIM: 0.8825 -> 0.8903 | Improved |
| replace SRN with SFARL | deconvolution with multiple degradations · Disk7 | PSNR: 26.46 -> 26.66 | Improved |
| replace GMM with SFARL | synthetic deraining · 12 images average | SSIM: 0.91 -> 0.95 | Improved |
| replace DDNET with SFARL | deraining · Rain1400 | PSNR: 29.91 -> 31.37 | Improved |
| replace DDNET with SFARL on unseen rain type | deraining · Rain100L | PSNR: 29.12 -> 29.73 | Improved |
Closest papers by meaning
- GIQE: Generic Image Quality Enhancement via Nth Order Iterative Degradation (CVPR 2022 · similarity 0.89)
- Rain Streak Removal Using Layer Priors (CVPR 2016 · similarity 0.89)
- Sparse Sampling Transformer with Uncertainty-Driven Ranking for Unified Removal of Raindrops and Rain Streaks (ICCV 2023 · similarity 0.89)
- Not Just Streaks: Towards Ground Truth for Single Image Deraining (ECCV 2022 · similarity 0.88)
- Joint Bi-Layer Optimization for Single-Image Rain Streak Removal (ICCV 2017 · similarity 0.88)
- MODEM: A Morton-Order Degradation Estimation Mechanism for Adverse Weather Image Recovery (NeurIPS 2025 · similarity 0.87)
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.
