On merging high- and low-resolution DEMs from TOPSAR and SRTM using a prediction-error filter

High-resolution digital elevation models (DEMs) are often limited in spatial coverage; they also may possess systematic artifacts when compared to comprehensive low-resolution maps. Here we correct artifacts and interpolate regions of missing data in airborne Topographic Synthetic Aperture Radar (TOPSAR) DEMs using a low-resolution Shuttle Radar Topography Mission (SRTM) DEM. We use a prediction error (PE) filter to interpolate and fill missing data so that the interpolated regions have the same spectral content as the valid regions of the TOPSAR DEM. The SRTM DEM is used as an additional constraint in the interpolation. We use cross-validation methods to obtain the optimal weighting for the PE filter and SRTM DEM constraints.