Resolution Enhanced SAR Tomography: From Match Filtering to Compressed Sensing Beamforming Techniques

For the analysis of forested scenes, the use of the sum of Kronecker products (SKP) decomposition technique allows the data covariance matrix of multipolarimetric multibaseline (MPMB) SAR surveys to be represented through a sum of two Kronecker products, composed of the polarimetric signatures and structural components for ground and canopy, respectively. Thus, different tomographic SAR focusing methods can be applied on the ground and canopy structural components separately. Nevertheless, the SKP decomposition may result in rank-deficient covariance matrices, restricting the usage of adaptive beamforming techniques. For this reason, this paper considers two robust beamforming approaches, the wavelet-based compressed sensing (CS) technique and the robust Capon beamforming (RCB) technique that incorporate additional constraints to guarantee robustness against rank deficiencies of the data covariance matrix. The reported experimental results, obtained with a real P-band MPMB data set (courtesy of the German Aerospace Center), reveal that the RCB method significantly out-performs the CS technique in the reduction of the computational complexity.