Research map / Image generation / Image restoration
HDNet: High-Resolution Dual-Domain Learning for Spectral Compressive Imaging
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use dual-domain supervision with FDL alongside SDL (ablation) | HDNet · KAIST test, 10 scenes (average) | PSNR: 31.91 -> 34.34 | Helped |
| Use dual-domain supervision with FDL alongside SDL (ablation) | HDNet · KAIST test, 10 scenes (average) | SSIM: 0.9093 -> 0.9572 | Helped |
| Set FDL loss weight to λ=0.7 rather than λ=0.9 (ablation) | HDNet · KAIST test, 10 scenes (average) | PSNR: 33.75 -> 34.34 | Helped |
| Increase FDL loss weight from λ=0.7 to λ=0.9 (ablation) | HDNet · KAIST test, 10 scenes (average) | PSNR: 34.34 -> 33.75 | Hurt |
| Set frequency-distance coefficient α=2 rather than α=3 (ablation) | HDNet · KAIST test | SSIM: 0.9065 -> 0.9572 | Helped |
| Set frequency-distance coefficient α=1 rather than α=2 (ablation) | HDNet · KAIST test | PSNR: 34.34 -> 34.40 | Helped |
| Use patch-based frequency supervision with p=3 rather than whole-image frequency supervision with p=1 (ablation) | HDNet · KAIST test | PSNR: 32.96 -> 34.34 | Helped |
| Increase patch subdivision from p=3 to p=4 (ablation) | HDNet · KAIST test | PSNR: 34.34 -> 33.69 | Hurt |
| Use high-resolution spatial-spectral attention instead of TSA-Net attention (ablation) | ResNet attention model · KAIST test, 10 scenes (average) | PSNR/SSIM: 32.68 / 0.9267 -> 34.34 / 0.9572 | Helped |
| Replace high-resolution attention with lower-resolution TSA attention (ablation) | HDNet spatial-spectral attention · HSI reconstruction attention visualization | high-frequency detail retention | Hurt |
| Remove frequency-domain supervision (ablation) | HDNet · HSI frequency reconstruction | LFD | Hurt |
Closest papers by meaning
- SPECAT: SPatial-spEctral Cumulative-Attention Transformer for High-Resolution Hyperspectral Image Reconstruction (CVPR 2024 · similarity 0.89)
- Dual-Window Multiscale Transformer for Hyperspectral Snapshot Compressive Imaging (AAAI 2024 · similarity 0.88)
- Learning Multi-Modal Cross-Scale Deformable Transformer Network for Unregistered Hyperspectral Image Super-resolution (AAAI 2024 · similarity 0.87)
- HIR-Diff: Unsupervised Hyperspectral Image Restoration Via Improved Diffusion Models (CVPR 2024 · similarity 0.87)
- EigenSR: Eigenimage-Bridged Pre-Trained RGB Learners for Single Hyperspectral Image Super-Resolution (AAAI 2025 · similarity 0.87)
- Improving Spectral Snapshot Reconstruction with Spectral-Spatial Rectification (CVPR 2024 · similarity 0.86)
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.