Determination of Microstructure Descriptors from Digital Images of Porous Media

Abstract

Preparation of polished sections of macroporous solids as well as their digital imaging, processing and analysis are described. The principal feature of the procedure is the image transformation to the corresponding phase function, which is a starting point in evaluation of statistical information on texture. This information can be used for three-dimensional stochastic reconstruction of porous solids. Out of many possible combinations of image processing methods, the following sequence is the most appropriate for our aim: (1) non-linear (median) filtration, (2) image segmentation based on maximization of grey-level variance between foreground and background objects, (3) removal of small clusters of pixels. As an additional criterion in optimization of computational parameters of the image processing methods, independent textural information was obtained by other experimental techniques, such as helium pycnometry, mercury porosimetry and low-temperature adsorption of krypton.