Parametric estimation in a model for PolSAR images

The scaled Complex Wishart distribution is a model that fits very well multilook full polarimetric Synthetic Aperture Radar (PolSAR) data from homogeneous regions. Its parameters, L and Sigma, are intrinsicaly related to the physical process of image aquisition giving information about the number of looks (L) and brightness (vertical bar Sigma vertical bar) of the image. Several techniques commonly used in image processing and understanding, such as image classification and segmentation, need to estimate these parameters. Assuming L known, in this work we propose a robust estimate of Sigma based on Huber ' s function. It is computed by means of a fixed point algorithm and its performance under different contamination scenarios is compared to that of the Maximun Likelihood estimator by means of MonteCarlo experiments. Stochastic distances and matrix bias are used to assess these performances. Results here obtained suggest the convenience of using a robust estimator of Sigma.