One of the challenges in converting a high-resolution voxelized image into a network is the difference in scales: millions or even billions of voxels must be mapped to a few thousand particles or pores, and we must ensure that this mapping process faithfully captures pore morphology at this coarser scale.

For particulate materials, network generation is performed in a two step process. The first is identification of individual particles. The second is network generation using the packing structure as a template. This approach has two advantages. The first is speed: creation of the particle map is fast and robust (i.e., relatively insensitive to image resolution). Subsequently, the network generation algorithm operates using a data set with thousands of elements (grains) rather than tens of millions of elements (voxels). The second advantage is that the packing structure and pore structure have the same characteristic scale. Hence, a pore network model that is created with the particle structure as a template has a better chance of capturing the fundamental pore morphology.

The grain-based algorithm is depicted in the above images for a quartz marine sand. The sample was supplied by A.H. Reed (NRL) and the imaging was performed by C.S. Willson (LSU) at the GSECARS beamline at Argonne National Laboratory..