The three-dimensional architecture of fibre network materials is characterised using X-ray computed tomography. While the generation of 3D reconstructions from 2-D images by employing back-projection algorithms has become a fairly routine operation, quantitative architecture data can often be inaccurate due to the finite volume of resolution and the noise due to tomographic reconstruction or data acquisition. In our group, an iterative deconvolution technique is employed to enhance the resolution of the filtered back-projections, so as to facilitate segmentation by means of a spatially adaptive thresholding algorithm. To extract architecture characteristics such as the distributions of fibre orientation and fibre segment length, and the number of inter-fibre crossings, skeletonisation algorithms are employed to reduce the 3D reconstructed fibres to their medial axes (skeletons), preserving their topology and basic shape. A further aim is to capture the effects of complex architecture, based on the numerical models of three-dimensional reconstructions generated from tomography scans. To do this, the reconstructed volume has to be discretised to form a 3D solid mesh, so that finite-element (FE) can be formulated and solved for a prescribed set of boundary conditions.