Fig. 5
From: Multi-channel volume density neural radiance field for hyperspectral imaging
Comparison of output image quality between the two methods with different network widths. The top row contains images generated by our method, the middle row contains images generated by original hyperspectral NeRF. Each column in the above rows stands for the output of specific network width. The bottom row contains the ground truth images. There are more fuzzy areas caused by floating objects in images generated by single sigma method than those by ours. One example is the small floating object ahead of the yellow arrow in the left-bottom zoomed in image with yellow rectangles.
