Research map / Image generation / Super-resolution
GEWDiff: Geometric Enhanced Wavelet-based Diffusion Model for Hyperspectral Image Super-resolution
Sirui Wang, Jiang He, Natàlia Blasco Andreo, Xiao Xiang Zhu
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 PCA to the wavelet encoder (ablation) | GEWDiff · Validation set | PSNR: 15.788 -> 27.013 | Improved |
| add RWA to the encoder (ablation) | GEWDiff · Validation set | PSNR: 25.640 -> 27.013 | Improved |
| add edge-aware noise perturbation (ablation) | GEWDiff · Validation set | PSNR: 26.579 -> 27.013 | Improved |
| add mask conditioning (ablation) | GEWDiff · Validation set | SSIM: 0.6530 -> 0.6573 (PSNR 26.681 -> 27.013) | No clear effect |
| use multi-level loss (pixel+perceptual+gradient) instead of a single term (ablation) | GEWDiff · Validation set | PSNR: 26.101 (pixel only) -> 27.013 | Improved |
| use perceptual loss only instead of full multi-level loss (ablation) | GEWDiff · Validation set | PSNR: 27.013 -> 21.664 | Hurt |
| add spectral fidelity enhancer (SFE) (ablation) | GEWDiff · Validation set | PSNR: 26.388 -> 27.013 | Improved |
| use 3D U-Net instead of 2D U-Net (ablation) | GEWDiff · Validation set | FID: 65.985 -> 30.110 | Improved |
| use more noise on edge (variant J) instead of default edge scheduler (ablation) | GEWDiff · Validation set | PSNR: 26.8713 vs 27.013 | No clear effect |
| use GEWDiff instead of SNLSR on SAM (MDAS sample 1) | HSI SR · MDAS sample 1 | SAM: 7.8911 (SNLSR) vs 8.4283 (GEWDiff) | Hurt |
| use GEWDiff instead of best baseline | HSI SR · MDAS sample 1 | FID: 49.026 -> 44.464 | Improved |
| apply GEWDiff after HySure fusion on real EnMAP data | EnMAP+Sentinel-2 fusion · real-world application | MetaIQA: 0.1997 -> 0.2029 | Improved |
Closest papers by meaning
- Enhancing Hyperspectral Images via Diffusion Model and Group-Autoencoder Super-resolution Network (AAAI 2024 · similarity 0.92)
- HSR-Diff: Hyperspectral Image Super-Resolution via Conditional Diffusion Models (ICCV 2023 · similarity 0.90)
- S2CycleDiff: Spatial-Spectral-Bilateral Cycle-Diffusion Framework for Hyperspectral Image Super-resolution (AAAI 2024 · similarity 0.89)
- HIR-Diff: Unsupervised Hyperspectral Image Restoration Via Improved Diffusion Models (CVPR 2024 · similarity 0.88)
- EigenSR: Eigenimage-Bridged Pre-Trained RGB Learners for Single Hyperspectral Image Super-Resolution (AAAI 2025 · similarity 0.88)
- Unmixing Diffusion for Self-Supervised Hyperspectral Image Denoising (CVPR 2024 · 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.
