Research map / Image generation / Image restoration
Degradation-Agnostic Clarity Learning for Unpaired Image Dehazing
Zhiyuan Song, Hannan Lu, Haiqian Han, Bo Li, Chang Liu, Pengxu Wei, Xiangyang Ji, Liang Lin
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 adversarial degradation mining (ARL) (ablation) | CycleGAN-Turbo baseline · RTTS | HazDesNet: 0.372 -> 0.282 | Improved |
| add adversarial degradation mining (ARL) (ablation) | CycleGAN-Turbo baseline · RTTS | MUSIQ: 56.63 -> 53.38 | Hurt |
| add Pareto optimization on top of ARL (ablation) | DCL · RTTS | HazDesNet: 0.282 -> 0.264 | Improved |
| add Pareto optimization on top of ARL (ablation) | DCL · RTTS | MANIQA: 0.272 -> 0.243 | Hurt |
| add E2R texture enhancement on top of ARL+PO (ablation) | DCL · RTTS | CLIPIQA: 0.299 -> 0.325 | Improved |
| add E2R texture enhancement on top of ARL+PO (ablation) | DCL · RTTS | MUSIQ: 54.67 -> 57.47 | Improved |
| add E2R texture enhancement on top of ARL+PO (ablation) | DCL · RTTS | HazDesNet: 0.264 -> 0.278 | Hurt |
| add E2R texture enhancement (ablation) | DCL · FoggyDriving (Person_Public33) | HazDesNet: 0.2449 -> 0.1498 | Improved |
| add E2R texture enhancement (ablation) | DCL · FoggyDriving (Person_Public33) | MUSIQ: 49.62 -> 53.69 | Improved |
| increase pool size K from 1 to 3 for Pareto optimization (ablation) | DCL · visual (Fig. 8) | artifacts | Improved |
| increase pool size K from 3 to 5 (ablation) | DCL · visual (Fig. 8) | artifacts | Hurt |
| replace CycleGAN-Turbo baseline with DCL | CycleGAN-Turbo · Unpaired synthetic heavy-haze | PSNR: 13.76 -> 14.51 | Improved |
Closest papers by meaning
- ClearAIR: A Human-Visual-Perception-Inspired All-in-One Image Restoration (AAAI 2026 · similarity 0.89)
- Advancing Real-World Image Dehazing: Perspective, Modules, and Training (TPAMI 2024 · similarity 0.89)
- BD-Diff: Generative Diffusion Model for Image Deblurring on Unknown Domains with Blur-Decoupled Learning (IJCV 2026 · similarity 0.88)
- Vision-Language Gradient Descent-driven All-in-One Deep Unfolding Networks (CVPR 2025 · similarity 0.88)
- Exploiting Diffusion Prior for Real-World Image Dehazing with Unpaired Training (AAAI 2025 · similarity 0.88)
- Unsupervised Blind Image Deblurring Based on Self-Enhancement (CVPR 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.
