Fully polarimetric 3D-ISAR helps to achieve better three-dimensional reconstruction with respect to the single polarization 3D-ISAR approach. This is useful to obtain more populated point clouds and to improve the estimation of the interferometric phases of the scattering centres that belong to the target. A more reliable reconstruction improves the performance of ATR (Automatic Target Recognition) algorithms. The whole process relies upon the scattering centres extraction procedure, through which the complex amplitudes of the scatterers are estimated. By using polarimetry within this iterative procedure, better results can be obtained. More than one approach is available in the literature: the Polarimetric CLEAN searches for the channel where the brightest scatterer has the highest power, while the Modified Polarimetric CLEAN arranges the data (iteration by iteration) in a subspace where either the SNR is maximum or the interference is minimum.We analyzed the performance of the two techniques in terms of accuracy of polarimetric features estimation by using a simulated target. The results suggest that the Polarimetric CLEAN already performs remarkably well, but there is still space for improvement, and the Modified Polarimetric CLEAN can actually achieve better results.