This thesis deals with the extraction of 3D buildings from optical and SAR high resolution spaceborne remote sensing images. The geometrical effects, resulting from the projection of 3D buildings in optical and SAR images, are first analyzed. Within this study, the geometrical signature of a building in an image is defined. It refers to all geometrical features of the image resulting from the presence of the building and can equivalently be described by a set of regions or by a set of segments in the image. From this analysis, methods for the simulation of building geometrical signatures in both optical and SAR images are then designed. Afterwards, two methods, based on the previous building geometrical signature simulation methods, are proposed for the extraction of 3D buildings in optical and SAR monoscopic imagery. These methods, which performances are assessed on both simulated and real data, are compliant with the two following requirements: - the geometrical and radiometrical complexity of high and very high resolution remote sensing images in urban environment has to be taken into account; - the proposed approaches can be implemented in optical as well as in SAR imagery, so that they can easily be used in a multi sensor context.
