An efficient, Bloch wave-based method is presented for simulation of high-resolution scanning confocal electron microscopy (SCEM) images. The latter are predicted to have coherent nature, i.e. to exhibit atomic contrast reversals depending on the lens defocus settings and sample thickness. The optimal defocus settings are suggested and the 3D imaging capabilities of SCEM are analyzed in detail. In particular, by monitoring average image intensity as a function of the probe focus depth, it should be possible to accurately measure the depth of a heavy-atom layer embedded in a light-element matrix.