Improving estimation in speckled imagery (Q2488421)

The \(\mathcal G_I^0\) distribution is characterized by it's PDF \[ F(z)={ L^L\Gamma(L-\alpha)\over \gamma^\alpha\Gamma(-\alpha)\Gamma(L)} {z^L\over (\gamma+zL)^{L-\alpha}}\mathbf{1}\{z>0\}, \] where \(\alpha<0\), \(\gamma>0\), \(L\geq 1\) are parameters. The authors propose \(\mathcal G_I^0\) as a model for a speckle noise distribution in such systems as Synthetic Aperture Radar, sonar, laser imagery. Four procedures for \(\alpha\) and \(\gamma\) estimation (\(L\) known) are considered: the method of moments, the quantile estimator, the maximum likelihood and it's bias corrected version. Results of simulations are presented.