A hierarchical Markov random field for road network extraction and its application with optical and SAR data

In this paper, we propose a hierarchical Markovian framework to extract the road network with optical and synthetic aperture radar (SAR) data. We propose a generalization of a previous method based on a low-level step (features extraction) and a high-level step (use of contextual information). The main novelties of the proposed approach are the use of more general elements to represent road candidates, which simplifies and generalizes the method, the fusion of different sensors during both lower and higher levels and the introduction of a second MRF in a hierarchical way. The approach is tested and evaluated using TerraSAR-X and Quickbird data.