Research map / Image generation / Image restoration
D2HNet: Joint Denoising and Deblurring with Hierarchical Network for Robust Night Image Restoration
Yuzhi Zhao, Yongzhe Xu, Qiong Yan, Dingdong Yang, Xuehui Wang, Lai-Man Po
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 |
|---|---|---|---|
| add the long-exposure input to short-only input (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 34.08 -> 34.67 | Improved |
| add the short-exposure input to long-only input (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 24.99 -> 34.67 | Improved |
| use s_first instead of l_last as ground truth (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 30.16 -> 34.67 | Improved |
| use deformable convolution in Alignment blocks instead of ordinary convolution (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 34.43 -> 34.67 | Improved |
| add Alignment and Feature fusion blocks (feature-level short-cuts) to EnhanceNet (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 33.27 -> 34.67 | Improved |
| add tail Residual block to EnhanceNet (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 34.17 -> 34.67 | Improved |
| add EnhanceNet after DeblurNet (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 30.48 -> 34.67 | Improved |
| add VarmapSelection of blurry training patches (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 34.24 -> 34.67 | Improved |
| add Illumination Adjustment augmentation (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 34.31 -> 34.67 | Improved |
| add Color Adjustment augmentation (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 34.58 -> 34.67 | Improved |
| add CutNoise augmentation (ablation) | D2HNet · D2-Dataset 1440p val | PSNR: 34.42 -> 34.67 | Improved |
| use D2HNet instead of simple concatenated two-phase pipelines TP1-TP3 | D2HNet · D2-Dataset 1440p and 2880p val | PSNR: 34.41 -> 34.67 | Improved |
Closest papers by meaning
- VQCNIR: Clearer Night Image Restoration with Vector-Quantized Codebook (AAAI 2024 · similarity 0.87)
- DarkIR: Robust Low-Light Image Restoration (CVPR 2025 · similarity 0.87)
- Interpretable Unsupervised Joint Denoising and Enhancement for Real-World low-light Scenarios (ICLR 2025 · similarity 0.86)
- NightHaze: Nighttime Image Dehazing via Self-Prior Learning (AAAI 2025 · similarity 0.86)
- Contrastive Learning for Low-Light Raw Denoising (Student Abstract) (AAAI 2024 · similarity 0.85)
- BlurDM: A Blur Diffusion Model for Image Deblurring (NeurIPS 2025 · similarity 0.84)
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.
