Research map / Image generation / Fast diffusion sampling
GMODiff: One-Step Gain Map Refinement with Diffusion Priors for Efficient HDR Reconstruction
Tao Hu, Weiyu Zhou, Yanjie Tu, Wei Dong, Peng Wu, Qingsen Yan, Yanning 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 |
|---|---|---|---|
| train reconstruction branch jointly with segmentation head (DaReg-Seg) (ablation) | DaReg · benchmark test set | image quality | Hurt |
| add the degradation-aware decoder (DA Decoder) (ablation) | GMODiff · benchmark test set | image quality | Improved |
| remove other modules one at a time (ablation) | GMODiff · benchmark test set | image quality | No clear effect |
| add LDM refinement on top of DaReg regression (ablation) | DaReg · benchmark test set | perceptual vs distortion | Mixed |
| GMODiff versus DiffHDR on perceptual quality | DiffHDR · In-distribution test split | MUSIQ: 60.39 -> 60.96 | Improved |
| GMODiff versus DiffHDR | DiffHDR · In-distribution test split | PU21-PSNR: 35.43 -> 36.68 | Improved |
| GMODiff versus AFUNet on distortion | AFUNet · In-distribution test split | PU21-PSNR: 36.91 -> 36.68 | Hurt |
| GMODiff versus AFUNet on perceptual DISTS | AFUNet · In-distribution test split | DISTS: 0.0120 -> 0.0103 | Improved |
| GMODiff versus AFUNet on out-of-distribution | AFUNet · Prabhakar dataset | DISTS: 0.0589 -> 0.0379 | Improved |
| GMODiff versus AFUNet on out-of-distribution | AFUNet · Prabhakar dataset | PU21-PSNR: 27.78 -> 26.86 | Hurt |
| one-step denoising instead of multi-step diffusion | DiffHDR · Inference 1080x1920 | inference time: 13.52s -> 0.93s | Improved |
| one-step denoising versus prior LDM-based LE-Diff | LE-Diff · Inference 1080x1920 | inference time: 132.7s -> 0.93s | 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.
- LE-Diff — LEDiff: Latent Exposure Diffusion for HDR Generation (CVPR 2025) · better on 1 of 1 shared result
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
- Single Shot HDR Recovery via a Video Diffusion Prior (NeurIPS 2026 · similarity 0.91)
- Generating Content for HDR Deghosting from Frequency View (CVPR 2024 · similarity 0.91)
- Zero-Shot Structure-Preserving Diffusion Model for High Dynamic Range Tone Mapping (CVPR 2024 · similarity 0.88)
- Diffusion Priors for Variational Likelihood Estimation and Image Denoising (NeurIPS 2024 · similarity 0.87)
- LEDiff: Latent Exposure Diffusion for HDR Generation (CVPR 2025 · similarity 0.87)
- GaussHDR: High Dynamic Range Gaussian Splatting via Learning Unified 3D and 2D Local Tone Mapping (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.
