I was reading this paper, and in there (3rd page, 18th line from the bottom) the authors have mentioned:
"we calculate ... the first 7 invariant Hu moments over the intensity values of voxels within the candidate mask." (A candidate mask is a 3d mask for a particular lung nodule/false positive candidate in a CT image.)
Everyone around me is saying the Hu's moments are only defined for 2D images. A quick search through literature seems to back them up. So, can anyone guess what the authors are probably doing here? (or Hu's moments can actually be extended to 3D?)