Research Article
Superpixel Spectral Unmixing for Hyperspectral Image Superresolution Using a Coupled Encoder-Decoder Network
Table 4
The average RMSE, PSNR, SAM, and ERGAS over CAVE and Harvard datasets.
| Methods | CNMF [8] | GSOMP+ [4] | BSR [24] | SNNMF [9] | CSU [10] | NSSR [5] | uSDN [21] | Ours |
| CAVE | RMSE | 8.98 | 8.05 | 4.93 | 8.45 | 6.73 | 4.67 | 4.12 | 3.69 | PSNR | 28.72 | 30.42 | 35.86 | 29.33 | 32.40 | 36.65 | 37.48 | 38.02 | SAM | 11.55 | 14.47 | 7.54 | 9.75 | 8.17 | 7.26 | 6.93 | 6.53 | ERGAS | 0.79 | 0.82 | 0.62 | 0.73 | 0.68 | 0.58 | 0.51 | 0.46 |
| Harvard | RMSE | 3.98 | 4.18 | 2.47 | 3.48 | 3.07 | 2.29 | 1.78 | 1.68 | PSNR | 38.11 | 37.35 | 42.36 | 40.03 | 41.07 | 42.63 | 44.22 | 45.02 | SAM | 4.87 | 5.02 | 3.43 | 4.25 | 3.83 | 3.31 | 2.85 | 2.63 | ERGAS | 0.43 | 0.45 | 0.31 | 0.39 | 0.36 | 0.28 | 0.22 | 0.19 |
|
|