It was assumed that each measurement contains values from at most two materials. Two-material mixtures were chosen based on a dimensionality argument. In an object that consists of regions of pure materials, as shown in Fig. 6, voxels containing one material will be most prevalent because they correspond to volumes. Voxels containing two materials will be next most prevalent, because they correspond to surfaces where two materials meet. As such, they are the first choice to model after those containing a single material. The approach can be extended in a straightforward manner to handle the three-material case as well as cases with other less frequent geometries, such as skin, tubes, or points where four materials meet. This extension could be useful for identifying subvoxel-sized geometry within sampled data, thus extending the resolution.
Was this article helpful?