Optimal removal of topological artefacts in microtomographic images of porous materials

Abstract

The analysis of flow at the pore scale in porous media has been facilitated with the use of microtomography. A powerful tool for quantifying the fluid structure using these tomographic 3D reconstructions is skeletonisation, but the significant disadvantage of this method is its sensitivity to noise, resulting in artefacts in the skeleton. A pre-processing of the 3D image is therefore required, but no method has yet proven to completely solve this problem. By developing a new procedure that, by construction, directly identifies the voxels and only those that are responsible for topological artefacts in the skeleton, we are able to remove all artefacts, and furthermore can prove that we do so by modifying a minimal amount of voxels in the segmented 3D image (i.e. the tomographic image in which each voxel has been assigned to either the porous or the solid phase). This is possible by identifying the three fundamental types of artefacts that can arise in a 3D skeleton, and dealing with each appropriately. Application to a microtomographic image of a sintered glass powder is presented. Impact of the different processing methods on the flow within its porosity is measured through the computed permeability deviations.