The addition of polarization diversity for a non-imaging laser speckle system provides improvement for the phase retrieval problem. The polarization diversity of the remote scene provides additional information for successfully recovering a two-dimensional image from noisy autocorrelations obtained from laser speckle patterns or pupil plane images. The proposed system may be used to characterize space-borne objects and debris with an earthbased sensor array. We propose an Expectation-Maximization (EM) algorithm with a simple, statistical-based stopping criteria. Results from both simulation and laboratory experiment are presented.