NASA, NSF and USGS jointly conducted airborne LIDAR surveys to acquire numerous surface points with high densities over the Antarctic Dry Valleys and its vicinity, The huge set of these points retains two characteristics undesirable for DEM generation, which are unusually high blunder ratio and large variation of the local point densities. Hence, in order to not only reduce the undesirable effects due to these characteristics but also process the huge number of points within reasonable limits of time and resources, we developed an efficient, robust, nearly automatic approach to DEM generation. This paper reports about the application of this approach to generating high-resolution precise DEMs from the Antarctic LIDAR surveys and the evaluation of their accuracy.