Many recent interferometric and tomographic SAR algorithms aimed at enhancing the monitoring performances on distributed areas affected by decorrelation phenomena have been proposed in the last years. These algorithms are based on the exploitation of the data covariance matrix, estimated on real data through the coherent averaging of statistically similar pixels, to improve signal-to-noise ratio and extract a decorrelation-filtered interferometric signal. Estimation of the covariance matrix on real data is, however, a challenging task since multilooking typically implies biased estimations as well as resolution losses which have to be properly handled. In this paper we discuss about the state-of-the-art and open issues for accurate estimation of covariance matrices on real data, for the applications in scatterer detection in SAR Tomography.