Research map / Image generation / Image restoration
Learning To Remove Refractive Distortions From Underwater Images
Simron Thapa, Nianyi Li, Jinwei Ye
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 distortion-guided GAN after distortion estimation (ablation) | Dis-Net · SynSet | PSNR: 20.015 -> 24.069 | Improved |
| add consistency loss (ablation) | Dis-Net · SynSet | PSNR: 18.090 -> 20.015 | Improved |
| add refraction loss (ablation) | Dis-Net · SynSet | PSNR: 17.853 -> 20.015 | Improved |
| add physics-based terms to distortion map loss (ablation) | Dis-Net · SynSet | PSNR: 15.010 -> 20.015 | Improved |
| add physics-based terms to distortion map loss (ablation) | Dis-Net · SynSet | SSIM: 0.358 -> 0.610 | Improved |
| add recurrent layers and three-frame input (temporal consistency) (ablation) | DG-Net-S · SynSet | distortion map RMSE: 0.0872 -> 0.0624 | Improved |
| add recurrent layers and three-frame input (temporal consistency) (ablation) | DG-Net-S · SynSet | PSNR: 21.198 -> 24.069 | Improved |
| use physics-guided DG-Net single input instead of Li et al. | DG-Net-S · SynSet | distortion map RMSE: 0.1089 -> 0.0872 | Improved |
| use DG-Net instead of Li et al. | DG-Net · SynSet | SSIM: 0.425 -> 0.800 | Improved |
| use DG-Net with 3 frames instead of James with 61 frames | DG-Net · ThapaSet | PSNR: 26.778 -> 26.021 | Hurt |
| use DG-Net with 3 frames instead of Oreifej with 61 frames | DG-Net · TianSet | PSNR: 20.457 -> 19.586 | Hurt |
| use DG-Net with 3 frames instead of Oreifej with 61 frames | DG-Net · TianSet | SSIM: 0.820 -> 0.840 | Improved |
Closest papers by meaning
- Unsupervised Underwater Image Restoration: From a Homology Perspective (AAAI 2022 · similarity 0.88)
- Unsupervised Non-Rigid Image Distortion Removal via Grid Deformation (ICCV 2021 · similarity 0.87)
- DRM-Net: Explicit Residual Modelling with Subaquatic Multi-Scale Context Fusion for Underwater Image Enhancement (AAAI 2026 · similarity 0.87)
- Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion (CVPR 2024 · similarity 0.85)
- Looking Into the Water by Unsupervised Learning of the Surface Shape (NeurIPS 2025 · similarity 0.85)
- Sea-ing in Low-light (CVPR 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.
