Research map / Image generation / Image restoration
FUMO: Prior-Modulated Diffusion for Single Image Reflection Removal
Telang Xu, Chaoyang Zhang, Guangtao Zhai, Xiaohong Liu
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 gated modulation (vs no gate) (ablation) | FUMO · Average (540) | PSNR: 26.55 -> 27.15 | Improved |
| add gated modulation (vs no gate) (ablation) | FUMO · Average (540) | MUSIQ: 57.13 -> 59.88 | Improved |
| add gated modulation (vs no gate) (ablation) | FUMO · Average (540) | LPIPS: 0.076 -> 0.069 | Improved |
| modulate residual injections with gate instead of concatenating priors (ablation) | FUMO · Average (540) | PSNR: 26.87 -> 27.15 | Improved |
| add high-frequency prior to intensity-only gate (ablation) | FUMO · Average (540) | CLIPIQA: 0.403 -> 0.424 | Improved |
| add intensity prior to HF-only gate (ablation) | FUMO · Average (540) | MUSIQ: 58.08 -> 59.88 | Improved |
| concatenate priors instead of no prior (no gate) (ablation) | FUMO · Average (540) | MUSIQ: 57.13 -> 58.55 | Improved |
| add fine-grained refinement module (vs coarse-only) (ablation) | FUMO · Average (540) | LPIPS: 0.133 -> 0.069 | Improved |
| add fine-grained refinement module (vs coarse-only) (ablation) | FUMO · Average (540) | SSIM: 0.865 -> 0.918 | Improved |
| add fine-grained refinement module (vs coarse-only) (ablation) | FUMO · Average (540) | PSNR: 26.27 -> 27.15 | Improved |
| use refinement module instead of fine-tuned decoder (ablation) | FUMO · Average (540) | PSNR: 26.75 -> 27.15 | Improved |
| fine-tuned decoder added to coarse-only (ablation) | FUMO · Average (540) | LPIPS: 0.133 -> 0.084 | Improved |
Closest papers by meaning
- RAP-SR: RestorAtion Prior Enhancement in Diffusion Models for Realistic Image Super-Resolution (AAAI 2025 · similarity 0.88)
- FAM Diffusion: Frequency and Attention Modulation for High-Resolution Image Generation with Stable Diffusion (CVPR 2025 · similarity 0.88)
- Learning Diffusion Texture Priors for Image Restoration (CVPR 2024 · similarity 0.87)
- A Modular Conditional Diffusion Framework for Image Reconstruction (NeurIPS 2024 · similarity 0.87)
- Revisiting Single Image Reflection Removal In the Wild (CVPR 2024 · similarity 0.87)
- FaithDiff: Unleashing Diffusion Priors for Faithful Image Super-resolution (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.
