Research map / Image generation / Image restoration
LAP-Net: Level-Aware Progressive Network for Image Dehazing
Yunan Li, Qiguang Miao, Wanli Ouyang, Zhenxin Ma, Huijuan Fang, Chao Dong, Yining Quan
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 multi-stage coarse-to-fine network instead of single stage (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: SSIM +0.051, CIEDE2000 0.88 better (0.678 -> 0.729) | Improved |
| add progressive level-aware supervision with average fusion over coarse-to-fine (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: 0.026 and 1.325 (0.729 -> 0.755) | Improved |
| add LSTM-generated stage weights over average fusion (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: 0.014 and 0.961 (0.755 -> 0.769) | Improved |
| add content-based (local entropy) weighting over average fusion (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: 0.032 and 1.356 (0.755 -> 0.787) | Improved |
| add contextual weighting on top of content-based weighting (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: 0.787 -> 0.798 SSIM, 14.640 -> 13.743 CIEDE2000 | Improved |
| haze-level guidance instead of sequential balance only for contextual weighting (ablation) | LAP-Net · O-HAZE | SSIM: 0.782 vs 0.784 | No clear effect |
| use contextual weighting without content-based weighting (ablation) | LAP-Net · O-HAZE | SSIM: not as good as content-based (0.787) | Hurt |
| replace top 0.1% dark-channel atmospheric light with A-subnet regression (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: 0.046 and 1.24 (0.735 -> 0.781) | Improved |
| treat atmospheric light as classification instead of regression (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: 0.017 and 0.75 (0.781 -> 0.798) | Improved |
| add Lc to Lt+LA loss (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: SSIM 0.791 -> 0.798, CIEDE2000 14.453 -> 13.743 | Mixed |
| Lt+LA only versus Lc only (ablation) | LAP-Net · O-HAZE | SSIM and CIEDE2000: SSIM 0.791 vs 0.777, CIEDE2000 14.453 vs 14.370 | Mixed |
Closest papers by meaning
- Proximal Dehaze-Net: A Prior Learning-Based Deep Network for Single Image Dehazing (ECCV 2018 · similarity 0.91)
- PHATNet: A Physics-guided Haze Transfer Network for Domain-adaptive Real-world Image Dehazing (ICCV 2025 · similarity 0.90)
- Adaptive Dynamic Dehazing via Instruction-Driven and Task-Feedback Closed-Loop Optimization for Diverse Downstream Task Adaptation (AAAI 2026 · similarity 0.88)
- PromptHaze: Prompting Real-world Dehazing via Depth Anything Model (AAAI 2025 · similarity 0.88)
- Depth Information Assisted Collaborative Mutual Promotion Network for Single Image Dehazing (CVPR 2024 · similarity 0.88)
- Dehaze-RetinexGAN: Real-World Image Dehazing via Retinex-based Generative Adversarial Network (AAAI 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.
