Research map / Image generation / Image restoration
How To Train Neural Networks for Flare Removal
Yicheng Wu, Qiurui He, Tianfan Xue, Rahul Garg, Jiawen Chen, Ashok Veeraraghavan, Jonathan T. Barron
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 flare loss L_F (ablation) | Ours + U-Net · Real flare test | PSNR: 24.84 -> 25.55 | Improved |
| add the flare loss L_F (ablation) | Ours + U-Net · Real flare test | SSIM: 0.841 -> 0.850 | Improved |
| add simulated (wave optics) scattering flare data (ablation) | Ours + U-Net · Real flare test | PSNR: 24.44 -> 25.55 | Improved |
| add simulated scattering flare data (ablation) | Ours + U-Net · Real flare test | SSIM: 0.843 -> 0.850 | Improved |
| add captured reflective flare data (ablation) | Ours + U-Net · Real flare test | PSNR: 23.77 -> 25.55 | Improved |
| add captured reflective flare data (ablation) | Ours + U-Net · Real flare test | SSIM: 0.828 -> 0.850 | Improved |
| use U-Net instead of CAN architecture (ablation) | Ours · Real flare test | PSNR: 24.21 -> 25.55 | Improved |
| use U-Net instead of CAN architecture (ablation) | Ours · Synthetic flare test | PSNR: 28.49 -> 30.37 | Improved |
| train with semi-synthetic flare data vs input image | Ours + U-Net · Real flare test | PSNR: 18.57 -> 25.55 | Improved |
| apply hand-crafted flare spot removal | Flare spot removal · Real flare test | PSNR: 18.57 -> 18.53 | No clear effect |
| apply dehazing algorithm to flare images | Dehaze · Synthetic flare test | PSNR: 21.13 -> 18.32 | Hurt |
| predict flare at 512x512 and upsample for 2048x2048 input | Ours + U-Net · 2048x2048 inference | seconds: 8s to 0.55s | Improved |
Closest papers by meaning
- Improving Lens Flare Removal with General-Purpose Pipeline and Multiple Light Sources Recovery (ICCV 2023 · similarity 0.92)
- Image Lens Flare Removal Using Adversarial Curve Learning (TPAMI 2025 · similarity 0.91)
- Disentangle Nighttime Lens Flares: Self-supervised Generation-based Lens Flare Removal (AAAI 2025 · similarity 0.90)
- FlareX: A Physics-Informed Dataset for Lens Flare Removal via 2D Synthesis and 3D Rendering (NeurIPS 2025 · similarity 0.89)
- Language-Guided One-Step Diffusion Model for Nighttime Flare Removal (CVPR 2026 · similarity 0.86)
- FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Views (CVPR 2025 · similarity 0.83)
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.
