Rock art studies are facing major technical challenges for extensive documentation. Nowadays, recording is essentially obtained from time-consuming tracing and rubbing, techniques that also require a high level of expertise. Recent advances in 3D modelling of natural objects and computational treatment of the modelled surfaces may provide an alternative, and reduce the current documentation bottleneck. The aim of this study is to examine the extent to which such treatments can be applied. The case study presented here concerns the famous deer stones erected by ancient Mongolian nomad populations. The 3D acquisition workflow is based on structure-from-motion, a versatile photogrammetric technique, well adapted to various field conditions. From the 3D geometry of objects of interest, elevation raster maps are produced by projection on four sides of the stela. These digital elevation models are then tested using algorithms based on differential geometry, sky visibility and local morphology, the general principles of which are briefly exposed. All these approaches may be appropriate with essentially planar surfaces. However, in the case of irregular carved surfaces, such as those of deer stones, the most efficient algorithm appears to be positive openness. In favourable cases, the incisions can be automatically delineated, facilitating the final drawing. Results obtained at the end of the process are comparable to the best drawings available in the literature, and can also include archaeological information about rock surface conditions. The procedure considerably accelerates the workflow in comparison with traditional techniques, reduces the level of expertise required, and provides 3D models, which can easily be shared, or further analysed by morphometric methods, for instance.