Search

The development of Synthetic Aperture Radar (SAR) techniques has been ongoing for a few decades due to the continuous expansion of SAR technologies. However, the source of decorrelations among other noise artefacts constitutes one of the main limitations to all interferometric-based processes including Tomography (TomoSAR). As a consequence, the interpretation of generated height maps and 3D point clouds is challenging, particularly with the difficulty of obtaining accurate in-situ measurements for validation. To address this issue in urban areas, a few works proposed the use of a filtering approach in order to facilitate the application of TomoSAR inversion or detection methods. At our end, we propose an improved version of multi-baseline Goldstein-based filter through its application in the wavelet domain, on the one hand, and the adaptive estimation of the alpha parameter, on the other hand. The generated height maps and point clouds using a limited number of images acquired by TerraSAR-X sensor are evaluated qualitatively and quantitatively. Edge preservation index and standard deviation showed the effectiveness of the approach with respect to the filtering criteria, while the R-squared and detection rate illustrated its efficiency in terms of height estimation. Both assessments demonstrated the performance of the proposed denoising methodology. [ABSTRACT FROM AUTHOR]