PROJECT TITLE :

On log-cumulants of multilook polarimetric whitening filter for polarimetric SAR data

ABSTRACT:

Many multivariate statistical distributions have been derived using the well-known product model to stochastically model PolSAR data. One important factor in their utilisation is the estimation of their texture parameters. The tactic primarily based on the second-order matrix log-cumulants for multilook PolSAR statistical distributions results in estimators with low bias and variance properties. Recently, it's been shown that the method of multivariate fractional moment (MOMFM) has an even lower mean-sq. error (MSE) than the tactic of multivariate log-cumulants-primarily based estimators, which relies on fractional moments of the multilook polarimetric whitening filter (MPWF). However, it performs worse in G0-Wishart distribution especially the form parameter is small. In this study, the authors propose new estimation ways based mostly on log-cumulants of the MPWF. One methodology relies on the first-order log-cumulants of the MPWF, that is appropriate for K-distribution case; the opposite technique relies on the second-order log-cumulants of the MPWF, which is appropriate for G0-distribution case. The bias, variance, MSE and computational time of the new estimators are all less than the MOMFM estimator. Comparisons are created among proposed ways and a few in style methods using simulated and real PolSAR data. The results show that new estimators have outstanding performance among the favored estimators. It's value to mentioning that the proposed estimators can be easily derived for all commonly occurring multilook PolSAR distributions.