We present numerical analysis of four microCT images of a Bentheimer outcrop sandstone sample acquired with different resolution. On the base of statistical analysis we show that the porosity and poretoskeleton distribution can be reliably obtained on samples of about 150 voxels with resolution coarser than 3 micrometers. In contrast, the statistically representative volume to determine complex properties such as permeability, tortuosity, or geometrical properties of the pore space may exceed 500 voxels and cannot be estimated reliably from conventional CT-scans. To overcome this difficulty we apply truncated Gaussian field technique for statistical simulation of images. We show that this technique allows to reproduce images with given porosity, pore-to-skeleton distribution, and permeability. However, geometrical properties of the simulated images do not match those of original CT-scans, and require higher-order statistics.