A comparison with different methods of statistically homogeneous pixels identification for multitemporal SAR images

The identification of statistically homogeneous pixels (SHP) plays an important role in SAR interferometry (InSAR). SHP selection is the basis of distributed scatterer (DS) InSAR. Aiming at the problem of effective extraction of SHP in DS research, we have analyzed four commonly used methods, namely Baumgartner-Weiβ-Schindler (BWS) test, T test, generalized likelihood ratio test (GLRT), and Fast Statistical Uniform Pixel Selection (FaSHPS) algorithm. These four methods have been compared through experiments using simulated data and real data in this paper. The results show that T test performs better than BWS test, but worse than GLRT and FaSHPS methods. In addition, the FaSHPS algorithm based on interval estimation provides higher computational efficiency, and works better than GLRT for small stacks.