Table 3 Performance comparison of output images between the two methods (the multi-sigma NeRF proposed in this paper and the original single-sigma NeRF) trained on datasets with different channel numbers.
From: Multi-channel volume density neural radiance field for hyperspectral imaging
Channel num | PSNR-single sigma | PSNR-multi sigma | Difference |
|---|---|---|---|
3 | 31.83 | 31.64 | -0.5% |
4 | 33.02 | 33.32 | +0.9% |
8 | 32.64 | 33.57 | +2.8% |
14 | 31.83 | 33.49 | +5.2% |
17 | 29.00 | 32.15 | +10.8% |
20 | 28.95 | 31.41 | +8.8% |
29 | 29.22 | 31.81 | +8.4% |
34 | 29.37 | 30.56 | +4.4% |
