An improved SAR tomographic technique using contextual information

Abstract:
SAR Tomography (TomoSAR) exploit multi-pass acquisitions and allow generating 3-D images, which provide an esti-mation of the scatterers distribution along range, azimuth and elevation. Usually the estimated elevation distribution of the scetterers in each range-azimuth pixel do not dependent on the neighboring ones (local approaches). In an urban environment, relationships among the elevations of neighboring pixels can be considerd exploiting its peculiar geometry. In this paper, the height profile is locally approximated with a planar surface and the parameters are estimated starting from the pixel values in the same position and in a “cluster” of neighboring positions. A GLRT approach is proposed to detect the presence of the scatteres. Results on simulated and real data validate the proposed approach.