Segmentation of multi-phase X-ray computed tomography images

Abstract

X-ray computed tomography (CT) is useful for non-destructively visualising internal features of non-transparent objects. In addition, it provides quantitative information about the geometry and spatial distribution of an object’s constituent materials. However, X-ray CT images include blurs and noises. Herein, the authors focus on the partial volume effect that causes blurs depending on image resolution. Artificial materials (e.g., thin wire and threads) and natural materials (e.g., oil sand) were observed using a microfocus X-ray CT scanner. A maximum likelihood thresholding method considering the partial volume effect based on histogram data was applied for segmenting two- and three-phase X-ray CT images. Relative errors in measured cross-sectional areas of aluminium wires and nylon threads were evaluated for verifying the segmentation method. Occupancy ratios of each phase of oil sand to its bulk volume were quantified. The introduced segmentation method estimates the volume of each constituent of two- and three-phase porous materials with reasonable accuracy.