Construction of 3D Digital Model of a Rock Sample Based on FIB-SEM Data

Focused ion beam tomography (FIB-SEM) is one of the promising methods for studying materials microstructure. The result of the procedure is an image stack of sequential layers of a sample. The reconstruction of the 3D porous structure from FIB-SEM data is an important problem in oil and gas industry for the moment. This paper concerns alignment and segmentation of images with pore-back effect. We introduce a metrics for evaluation of alignment quality and present a new segmentation method based on a marker-controlled watershed. Matrix markers are found with the use of variance filter, and markers of pores - by thresholding and morphological half-gradient. Besides that, a new approach is employed for generation of more reliable synthetic ground truth data. Finally, we evaluate the proposed segmentation method numerically using the synthetic image and manually labelled real data. Our segmentation technique outperforms two existing algorithms