Enhancing building reconstruction by adaptive filtering of interferometric phases

The new generation of space borne sensors like TerraSAR-X and TanDEM-X is very attractive for 3D information extraction because of their high resolution and their possibility to achieve repeat-pass and since short single-pass data. Moreover, they can cover a huge area and especially their interferometric constellations allow recovering the height of objects like buildings. Nevertheless, the detection of buildings in interferometric phase data turns out to be difficult because of the high interferometric phase noise. In this paper, the focus is put on filtering the phase image at building location to enhance the building reconstruction. Different filtering methods are investigated, from the standard area filters to orientation filters, relying only on the image statistics or taking into account GIS information. These specialized filters are tested on real single-pass airborne InSAR data and compared to standard filtering methods.