Research map / Image generation / Image restoration
Non-local Meets Global: An Integrated Paradigm for Hyperspectral Image Restoration
Wei Ping He, Quanming Yao, Chao Li, Naoto Yokoya, Qibin Zhao, Hongyan Zhang, Liangpei 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 |
|---|---|---|---|
| increase spectral dimension K linearly each iteration (delta=2 vs 0) (ablation) | NGmeet · PaC denoising (sigma^2 100) | PSNR (dB): 29.91 -> 30.83 | Improved |
| update K with delta=1 vs fixed K at low noise (ablation) | NGmeet · PaC denoising (sigma^2 10) | PSNR (dB): 43.09 -> 43.52 | Improved |
| raise step size from delta=2 to delta=3 (ablation) | NGmeet · PaC denoising (sigma^2 100) | PSNR (dB): 30.83 -> 30.45 | Hurt |
| re-match non-local similar groups each iteration (ablation) | NGmeet · denoising PSNR curves | PSNR | Improved |
| initialize K with HySime (ablation) | NGmeet · PaC denoising | PSNR | Improved |
| use WSNM instead of WNNM as patch denoiser (ablation) | NGmeet · PaU denoising (sigma 10) | PSNR (dB): 43.17 -> 43.44 | Improved |
| combine spectral low-rank with non-local similarity (NGmeet vs LLRT) | NGmeet · CAVE denoising (sigma 100) | PSNR (dB): 36.25 -> 37.19 | Improved |
| combine spectral low-rank with non-local similarity (NGmeet vs Fast-HyDe) | NGmeet · WDC denoising (sigma 100) | PSNR (dB): 31.05 -> 31.48 | Improved |
| non-local LLRT versus spectral FastHyDe on MSI versus HSI | LLRT / Fast-HyDe · CAVE vs PaC denoising | PSNR (dB): 46.72 -> 47.14 | Mixed |
| use NGmeet with stage A spectral projection instead of non-local KBR/LLRT only | NGmeet · CAVE denoising runtime | time (s): 1212 -> 204 | Improved |
| use NGmeet over the best prior method for compressive imaging | NGmeet · CAVE Toy CASSI | PSNR (dB): 25.91 -> 27.13 | Improved |
| use NGmeet for inpainting versus NLLRTC | NGmeet · CAVE inpainting (SR 5%) | PSNR (dB): 36.80 -> 38.10 | Improved |
Closest papers by meaning
- Non-Local Meets Global: An Integrated Paradigm for Hyperspectral Denoising (CVPR 2019 · similarity 0.96)
- Hyper-Laplacian Regularized Unidirectional Low-Rank Tensor Recovery for Multispectral Image Denoising (CVPR 2017 · similarity 0.92)
- Computational Hyperspectral Imaging Based on Dimension-Discriminative Low-Rank Tensor Recovery (ICCV 2019 · similarity 0.91)
- HIR-Diff: Unsupervised Hyperspectral Image Restoration Via Improved Diffusion Models (CVPR 2024 · similarity 0.90)
- Hyperspectral Image Super-Resolution via Non-Local Sparse Tensor Factorization (CVPR 2017 · similarity 0.90)
- 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.
