Research map / 3D vision / Gaussian splatting
DehazeGS: Seeing Through Fog with 3D Gaussian Splatting
Jinze Yu, Yiqun Wang, Aiheng Jiang, Zhengda Lu, Jianwei Guo, Yong Li, Hongxing Qin, Xiaopeng Zhang
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 |
|---|---|---|---|
| model fog with atmospheric scattering model on Gaussians (Lrec only) vs vanilla 3DGS (ablation) | 3DGS · Synthetic foggy Mip-NeRF scenes | PSNR: 12.27 -> 19.35 | Improved |
| add dark channel prior loss (ablation) | DehazeGS (Lrec) · Synthetic foggy Mip-NeRF scenes | PSNR / LPIPS: 19.35 -> 19.66 PSNR; 0.189 -> 0.173 LPIPS | Improved |
| add bright channel prior loss (ablation) | DehazeGS (Lrec+LDCP) · Synthetic foggy Mip-NeRF scenes | PSNR / SSIM / LPIPS: 19.66 -> 19.74 PSNR; 0.687 -> 0.697 SSIM; 0.173 -> 0.175 LPIPS | Mixed |
| add pseudo-depth regularization Ld (ablation) | DehazeGS (Lrec+LDCP+LBCP) · Synthetic foggy Mip-NeRF scenes | SSIM / LPIPS: 0.697 -> 0.692 SSIM; 0.175 -> 0.166 LPIPS; 19.74 -> 19.77 PSNR | Mixed |
| add depth-weighted reconstruction loss (ablation) | DehazeGS (Lrec+LDCP+LBCP+Ld) · Synthetic foggy Mip-NeRF scenes | PSNR / SSIM / LPIPS: 19.77 -> 19.83 PSNR; 0.692 -> 0.702 SSIM; 0.166 -> 0.164 LPIPS | Improved |
| dehaze images with Dehamer before 3DGS vs 3DGS on foggy images | 3DGS · Real foggy dataset | SSIM: 0.570 -> 0.423 | Hurt |
| DehazeGS vs DehazeNeRF on Lion scene | DehazeGS · Real foggy dataset, Lion scene | PSNR: 18.01 -> 17.33 | Hurt |
| DehazeGS vs SeaSplat on Counter scene | DehazeGS · Synthetic foggy, Counter scene | SSIM: 0.736 -> 0.704 | Hurt |
| DehazeGS vs DehazeNeRF on real data (average) | DehazeNeRF · Real foggy dataset | PSNR: 16.93 -> 17.77 | Improved |
Closest papers by meaning
- GaussianPro: 3D Gaussian Splatting with Progressive Propagation (ICML 2024 · similarity 0.88)
- HybridGS: Decoupling Transients and Statics with 2D and 3D Gaussian Splatting (CVPR 2025 · similarity 0.87)
- SynFog: A Photo-realistic Synthetic Fog Dataset based on End-to-end Imaging Simulation for Advancing Real-World Defogging in Autonomous Driving (CVPR 2024 · similarity 0.87)
- CoCoGaussian: Leveraging Circle of Confusion for Gaussian Splatting from Defocused Images (CVPR 2025 · similarity 0.87)
- IBGS: Image-Based Gaussian Splatting (NeurIPS 2025 · similarity 0.87)
- DropGaussian: Structural Regularization for Sparse-view Gaussian Splatting (CVPR 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.
