A two-stage approach to the segmentation of FIB-SEM images of highly porous materials

Segmentation of 3D FIB-SEM images is a specific problem for porous materials, where grey intensities are not sufficient to determine the phase represented by a certain voxel. In this paper a new approach is proposed to segmentation of FIB-SEM images, which is especially designed for highly porous materials. It detects and allocates structures based on their last occurrence in z-direction. Afterwards, the obtained segmentation is improved by applying local thresholds. This can be done in an iterative or in a direct way — both methods are described and compared with each other. Furthermore, the final segmentation is compared to a binarisation obtained by adaptive local thresholding.

► We describe a new algorithm for the segmentation of FIB-SEM images.
► Structures are detected by backpropagating local thresholds in z-direction.
► The algorithm is applied to a highly porous monolithic foam-like silica material.
► The result is compared to a segmentation obtained by local thresholding.