A new technique to quantify variations in pore structure across an image is described in this paper. The method involves several stages as the original images are rarely suitable for simple segmentation by the selection of a single threshold. The processing first involves the application of a Wiener filter to remove imaging artefacts and to sharpen the edge of void/particle contacts. This image reconstruction is followed by the selection of a global threshold based on the relative contrast method, to create a binary image separating the voids from the particles. Such an image can be used to evaluate pore structure and overall porosity, and this can be related to overall macroscopic porosity.
The method may be enhanced by combining the binary image with that from domain segmentation methods, where the image is automatically classified according to the orientation of fine grained materials. It is thus possible to study differences in the porosity of domains with vertically orientated particles from those with horizontal ones. The full procedure may be automated so that large numbers of images may be processed in a batch without operator intervention.
Some examples of the technique using back-scattered scanning electron microscope (SEM) images of polished blocks of embedded soil samples are given. As expected, the random areas of the sample have a consistently higher porosity than other areas, but contrary to expectations, the horizontal domains appear to be more porous than vertical ones.
Hounslow, M. W. and Tovey, N. K.
"Segmentation of Pores in Backscattered Images of Sediments and Soils and their Relationship to Domain Structure,"
Scanning Microscopy: Vol. 1992
, Article 23.
Available at: https://digitalcommons.usu.edu/microscopy/vol1992/iss6/23