Abstract : This survey paper focus on recent researches whose goal is to optimize treatments on 3D meshes, thanks to a study of their surface features, and more precisely their roughness and saliency. Applications like watermarking or lossy compression can benefit from a precise roughness detection, to better hide the watermarks or quantize coarsely these areas, without altering visually the shape. Despite investigations on scale dependence leading to multi-scale approaches, an accurate roughness or pattern characterization is still lacking, but challenging for those treatments. We think there is still room for investigations that could benefit from the power of the wavelet analysis or the fractal models. Furthermore only few works are now able to differentiate roughness from saliency, though it is essential for faithfully simplifying or denoising a 3D mesh. Hence we have investigated roughness quantification methods for analog surfaces, in several domains of physics. Some roughness parameters used in these fields and the additionnal information they bring are finally studied, since we think an adaptation for 3D meshes could be beneficial.