DInSAR Pixel Selection Based on Sublook Spectral Correlation Along Time

Abstract

The reliability of any final differential synthetic aperture radar (SAR) interferometry (DInSAR) product is directly related to the phase quality of the interferograms involved in the processing. Performing an appropriate selection of high-quality pixels is hence mandatory in order to avoid the inclusion of noisy data in the processing and thus obtain reliable deformation information within the area of study. In this framework, a new full-resolution pixel selection method based on exploiting the spectral properties of pointlike scatterers is presented. The method is developed from the concept of the so-called coherent scatterers, which are characterized by having a correlated spectrum in a single acquisition, but now involving the temporal axis in the detection. Using the spectral properties of scatterers to perform the DInSAR pixel selection step presents some advantages with respect to the well-known permanent scatterer (PS) approach. On the one hand, the radiometric calibration of data is not required since the detection is now not dependent on the amplitude behavior of scatterers. On the other hand, a reliable pixel selection can be performed even with a reduced number of images due to the nature of the estimator. In contrast, its main limitation relies on the loss of range resolution involved in the sublook generation process. The method is more suited for urban scenarios where the density of temporally stable pointlike scatterers is generally higher. In this context, the final deformation maps obtained using the new method presented are compared, in terms of density and phase quality, with the ones obtained with the well-known traditional PS approach, showing a similar performance. Finally, since both methods characterize pointlike scatterers from different points of view, the combination of both techniques is proposed leading to a significant increase in the pixels' density and giving place to a meaningful improvement in the final DInSAR results.