Research map / Image generation / Image restoration
Structure-Guided Diffusion Models for High-Fidelity Portrait Shadow Removal
Wanchang Yu, Qing Zhang, Rongjia Zheng, Wei-Shi Zheng
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 structure guidance to the inpainting diffusion model (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | SSIM: 0.757 -> 0.830 | Improved |
| add structure guidance (RMSE) (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | RMSE: 23.598 -> 17.162 | Improved |
| use shadow-independent SE-Net structure instead of PDG predicted structure (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | SSIM: 0.779 -> 0.830 | Improved |
| use shadow-independent SE-Net structure instead of PDG predicted structure (RMSE) (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | RMSE: 19.963 -> 17.162 | Improved |
| train SE-Net on relighting-based synthesized data instead of PSM synthesis (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | SSIM: 0.785 -> 0.830 | Improved |
| train SE-Net on relighting-based synthesized data instead of PSM synthesis (LPIPS) (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | LPIPS: 0.086 -> 0.056 | Improved |
| add detail restoration diffusion model (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | SSIM: 0.814 -> 0.830 | Improved |
| add detail restoration diffusion model (shadow region LPIPS) (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | LPIPS (shadow region): 0.020 -> 0.011 | Improved |
| add detail restoration diffusion model (RMSE) (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | RMSE: 18.537 -> 17.162 | Improved |
| replace fine-tuned Stable Diffusion + ControlNet inpainting with the trained inpainting diffusion model (ablation) | Our portrait shadow removal pipeline · PSM portrait shadow dataset | structure fidelity (qualitative) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ShadowDiffusion — ShadowDiffusion: When Degradation Prior Meets Diffusion Model for Shadow Removal (CVPR 2023) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Detail-Preserving Latent Diffusion for Stable Shadow Removal (CVPR 2025 · similarity 0.90)
- Diff-Shadow: Global-guided Diffusion Model for Shadow Removal (AAAI 2025 · similarity 0.89)
- PHR-DIFF: Portrait Highlights Removal via Patch-aware Diffusion Model (AAAI 2025 · similarity 0.89)
- Recasting Regional Lighting for Shadow Removal (AAAI 2024 · similarity 0.87)
- Polarization Guided Mask-Free Shadow Removal (AAAI 2025 · similarity 0.87)
- SynthLight: Portrait Relighting with Diffusion Model by Learning to Re-render Synthetic Faces (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.
