Semi-Automated Multiphase Segmentation of 4D Micro-Computed Tomography Data of Porous Media.

Noninvasive imaging using micro X-ray computed tomography (CT) is increasingly being used in the study of porous media for applications such as oil recovery and contaminant remediation. The advancements made in the imaging technology have led to generation of a large amount of X-ray CT data of porous media. A meaningful and scientific analysis of the data, however, lags behind. A crucial step in the analysis is the segmentation of the data into constituent phases such as solid, liquid and air. Due to the presence of multiple phases in porous media, compounded by the lack of stable three-dimensional multiphase segmentation algorithms for porous media, it is still a common practice to use manual thresholding or binary thresholding iteratively. To overcome such drawbacks in addition to minimizing any operator inputs, we present a semi-automated three-dimensional multiphase segmentation algorithm based on the K-means algorithm and Markov random field framework (KM-MRF). Furthermore, the KM-MRF and K-means methods were evaluated on two 4D datasets, each representing a time-series of 3D image data. The results for grayscale datasets of glass beads and 45-50 silica soil show a need for an advanced algorithm such as KM-MRF instead of simpler multiphase algorithms such as K-means. We also demonstrate the need for post-segmentation processing methods such as connected-component merging and morphological operations. Although the presented results are promising, further potential improvements for fully automated operations are discussed.