Table 4 Reconstruction accuracy comparison of representative SR methods in terms of RMSE, MRAE and SAM on the BGU-HS and ARAD-HS datasets. Top two best results are highlighted in bold and underline respectively.
From: A survey on computational spectral reconstruction methods from RGB to hyperspectral imaging
Category | Method | BGU-HS | ARAD-HS | ||||
|---|---|---|---|---|---|---|---|
RMSE | MRAE | SAM | RMSE | MRAE | SAM | ||
Dictionary learning | Sparse coding | 51.48 | 0.0808 | 5.01 | 0.0331 | 0.0787 | 6.46 |
SR A+ | 26.09 | 0.0448 | 2.83 | 0.0226 | 0.0725 | 4.61 | |
Linear CNN | HSCNN | 17.006 | 0.0190 | – | – | – | – |
SR-2DNet | 21.394 | 0.020 | – | – | – | – | |
SR-3Dnet | 20.010 | 0.018 | – | – | – | – | |
U-Net | SRUNet | 15.88 | 0.0156 | 1.11 | 0.0152 | 0.0395 | 2.74 |
SRMSCNN | 19.28 | 0.0231 | 1.47 | 0.0235 | 0.0724 | 4.91 | |
SRMXRUNet | – | – | – | – | 0.0454 | – | |
SRBFWU-Net | – | – | – | 0.0151 | 0.0434 | – | |
Dense network | SRTiramisuNet | 20.98 | 0.0272 | 1.57 | 0.0251 | 0.0850 | 4.34 |
HSCNN-R | 13.911 | 0.0145 | 1.05 | 0.0143 | 0.0372 | 2.63 | |
HSCNN-D | 13.128 | 0.0135 | 0.99 | – | – | – | |
Attention network | SRAWAN | 10.24 | 0.0114 | – | 0.0111 | 0.0312 | 2.16 |
SRHRNet | – | – | 1.01 | 0.0135 | 0.0423 | 2.53 | |
