A technique is described that has been used to semi-automatically generate Digital Terrain Models (DTMs) of the lunar surface using stereo pair images obtained from the Apollo Metric cameras, and from the Clementine spacecraft. This involves using digital images of the Moon's surface (scanned from photographs in the case of Metric) for automated stereo matching with the "Gotcha" software developed at University College London. The resulting stero disparity data are then passed through a camera model and, with the aid of some manually derived control points, a DTM can be produced. This entire process can be performed in under one day on a Sun Sparc 1 for a Metric stereo pair. Apollo Metric stero imagery has a ground resolution of 25m and a height precision of 30m, but is limited in coverage to near equatorial regions. Clementine stereo imagery has aground pixel size of 100m at best, and a height precision of the order of 1km over the majority of the Moon. However in the polar areas, and other regions that were imaged specifically for stereo, the height precision is improved to the order of 100m. The heigth imformation derived from the DTM yields important additional data for the geologic interpretation of surface features like: depth to diameter ratio and volume studies for craters down to a diameter of 1 km, the thickness of lava flows and crater ejecta, and cross-sections through tectonic features and impact basin rings. In addition the DTM allows us to enhance the morphology of surface features that normally appears camouflaged by albedo markings or shading, and they can provide local slope information to support lunar surface photometric studies. Examples are given of a DTM covering the region to the west of Tiolkovskiy from Apollo Metric, and experiments to stereo match the polar regions of the Moon from Clementine imagery.