High spatial resolution Digital Elevation Models (DEMs) are fundamental datasets to understand the dynamics of Earth surface, such as recent change of the cryosphere, and especially in remote mountainous areas, like the Andes. Here we describe a methodological framework to improve the quality in terms of representativeness of terrain surface for high spatial resolution stereoscopic DEM obtained from Pléiades images. Using as example two DEMs from highly different regions from the Andes, we analyzed the different types of errors present in both DEMs. In order to improve the output DEM a post-processing scheme was designed, following those steps: (1) filling the gaps with a spline interpolation method; (2) elimination of the granular noise with a multidirectional Lee filter; and (3) elimination of the spikes with a slope-based DTM filter. Besides improving the representativeness of both DEMs in terms of terrain feature, the post-processing also improves the accuracy of DEMs in terms of absolute accuracy of elevations.