Detailed analysis of scanning probe microscopy (SPM) data acquired for faceted and nonflat surfaces is usually complicated due to the presence of a large number of surface areas tilted by large/variable angles relative to the scanning plane. As a consequence, standard methods of elimination of global or local slopes by either plane subtraction or numerical differentiation seem to be ineffective. We demonstrate that a simple difference-of-Gaussians procedure provides output data corresponding to the projection of a considered surface onto the scanning plane without undesirable contrast modifications. This method allows us to suppress small-scale noise, minimize effects of finite slopes in the SPM images along both fast and slow scanning directions, and remove surface ripples without active participation of the operator. This method can be applied for fast on-the-fly visualization of experimental data and for more detailed analysis, including highprecision determination of lattice parameters and angles between translation vectors for surface reconstruction of different terraces or surface domains. In order to estimate geometrical distortions introduced by our procedure, we compare the results obtained by the difference-of-Gaussians approach for the tilted surfaces with direct image rotation in 3D space.