Markov random fields for digital terrain model extraction

This paper deals with the automatic extraction of the ground elevation or digital terrain model (DTM) from the global digital elevation model (DEM). The proposed method is divided into two main steps: first, the road network and the cross-roads are extracted, and the elevation of the cross-roads is estimated on the DEM; secondly, the height of the cross-roads is regularized using some contextual knowledge. The second step is performed inside a Markovian framework based on the natural graph of the roads. Results on interferometric SAR data and optical images are presented.