Research map / Image generation / Image restoration
Deep Variational Network Toward Blind Image Restoration
Zongsheng Yue, Hongwei Yong, Qian Zhao, Lei Zhang, Deyu Meng, Kenneth K. Wong
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 |
|---|---|---|---|
| condition image posterior on noise variance (ablation) | Baseline1 (q(z|y)) · DIV2K100 SR x4 (noise 2.55) | PSNR (dB): 27.69 -> 27.71 | No clear effect |
| condition image posterior on blur kernel parameters (ablation) | Baseline1 (q(z|y)) · DIV2K100 SR x4 (noise 2.55) | PSNR (dB): 27.69 -> 27.86 | Improved |
| condition image posterior on both noise and kernel (ablation) | Baseline3 (kernel only) · DIV2K100 SR x4 (noise 2.55) | LPIPS (lower is better): 0.3611 -> 0.3588 | Improved |
| condition image posterior on noise, in addition to kernel (ablation) | Baseline2 (noise only) · DIV2K100 SR x4 (noise 2.55) | PSNR (dB): 27.71 -> 27.86 | Improved |
| use VIRNet estimated noise level instead of ground truth as FFDNet input | FFDNet · CBSD68 AWGN (sigma=50) | PSNR (dB): 27.95 -> 27.92 | No clear effect |
| use VIRNet estimated noise level instead of ground truth as FFDNet input | FFDNet · CBSD68 AWGN (sigma=15) | PSNR (dB): 33.87 -> 33.87 | No clear effect |
| use VIRNet estimated noise map instead of ground truth as FFDNet input | FFDNet · CBSD68 non-i.i.d. noise (Case 1) | PSNR (dB): 28.79 -> 28.75 | No clear effect |
| use VIRNet estimated kernel instead of KernelGAN with USRNet | USRNet · DIV2K20 SR x4 | PSNR (dB): 14.52 -> 28.53 | Improved |
| use VIRNet estimated kernel instead of BSRDM with USRNet | USRNet · DIV2K20 SR x4 | PSNR (dB): 27.07 -> 28.53 | Improved |
| use VIRNet kernel instead of ground-truth kernel with USRNet | USRNet · DIV2K20 SR x4 | PSNR (dB): 29.19 -> 28.53 | Hurt |
| use blind VIRNet instead of non-blind DRUNet | DRUNet · CBSD68 non-i.i.d. noise (Case 2) | PSNR (dB): 28.64 -> 28.93 | Improved |
| use blind VIRNet instead of MPRNet | MPRNet · SIDD real-world denoising | PSNR (dB): 39.62 -> 39.63 | No clear effect |
Closest papers by meaning
- Variational Denoising Network: Toward Blind Noise Modeling and Removal (NeurIPS 2019 · similarity 0.92)
- DiffBIR: Toward Blind Image Restoration with Generative Diffusion Prior (ECCV 2024 · similarity 0.91)
- Taming Generative Diffusion Prior for Universal Blind Image Restoration (NeurIPS 2024 · similarity 0.89)
- Blind Image Restoration via Fast Diffusion Inversion (NeurIPS 2024 · similarity 0.88)
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.88)
- A Modular Conditional Diffusion Framework for Image Reconstruction (NeurIPS 2024 · 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.
