Research map / Image generation / Image restoration
Self-Diffusion Driven Blind Imaging
Yanlong Yang, Guanxiong Luo
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 DeblurSDI instead of best baseline (FFT-ReLU Deblur) | blind deblurring · Levin (optical aberration) | PSNR: 19.57 -> 28.36 | Improved |
| use DeblurSDI instead of best baseline (Phase-Only) | blind deblurring · FFHQ (optical aberration) | PSNR: 26.31 -> 33.00 | Improved |
| use DeblurSDI instead of SelfDeblur | blind motion deblurring · Levin (motion blur) | PSNR: 25.06 -> 31.85 | Improved |
| use DeblurSDI instead of Phase-Only | blind motion deblurring · Kohler (motion blur) | SSIM: 0.8092 -> 0.7653 | Mixed |
| use DeblurSDI instead of SelfDeblur | blind motion deblurring · Cho (motion blur) | PSNR: 20.37 -> 28.73 | Improved |
| evolve kernel latent with noise schedule (k-diff) instead of fixed latent (Standard) (ablation) | DeblurSDI kernel generator · blind deblurring (Figure 4) | PSNR/SSIM | Improved |
| deepen kernel generator (n=1 to n=5) (ablation) | DeblurSDI k-diff · blind deblurring (Figure 4) | kernel accuracy | Improved |
| increase outer diffusion steps T from 10 to 30 (ablation) | DeblurSDI · blind deblurring (Figure 2) | SSIM/PSNR | Improved |
| increase T from 30 to 40 or S beyond 400 (ablation) | DeblurSDI · blind deblurring (Figure 2) | SSIM/PSNR | No clear effect |
| raise additive noise sigma from 0.02 to 0.05 (ablation) | DeblurSDI · blind deblurring (Figure 5) | PSNR: 29.64 -> 20.48 | Hurt |
Closest papers by meaning
- Kernel Diffusion: An Alternate Approach to Blind Deconvolution (ECCV 2024 · similarity 0.88)
- Parallel Diffusion Models of Operator and Image for Blind Inverse Problems (CVPR 2023 · similarity 0.88)
- BlurDM: A Blur Diffusion Model for Image Deblurring (NeurIPS 2025 · similarity 0.87)
- A Diffusion Model with State Estimation for Degradation-Blind Inverse Imaging (AAAI 2024 · similarity 0.87)
- A Modular Conditional Diffusion Framework for Image Reconstruction (NeurIPS 2024 · similarity 0.86)
- Blind Image Restoration via Fast Diffusion Inversion (NeurIPS 2024 · similarity 0.86)
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.
