We present a process based method for reconstructing the full three-dimensional microstructure of sandstones. The method utilizes petrographical information obtained from two-dimensional thin sections to stochastically model the results of the main sandstone forming processes - sedimentation, compaction, and diagenesis. We apply the method to generate Fontainebleau sandstone and compare quantitatively the reconstructed microstructure with microtomographic images of the actual sandstone. The comparison shows that the process based reconstruction reproduces adequately important intrinsic properties of the actual sandstone, such as the degree of connectivity, the specific internal surface, and the two-point correlation function. A statistical reconstruction of Fontainebleau sandstone that matches the porosity and two-point correlation function of the microtomography data differs strongly from the actual sandstone in its connectivity properties. Transport properties of the samples are determined by solving numerically the local equations governing the transport. Computed permeabilities and formation factors of process based reconstructions of Fontainebleau sandstone compare well with experimental measurements over a wide range of porosity.