Maximum Likelihood texture tracking in highly heterogeneous PolSAR clutter

This paper introduces a generalisation of the conventional Maximum Likelihood (ML) texture tracking algorithm in the context of highly heterogeneous PolSAR clutter. The statistical criterion is defined in both uncorrelated and correlated texture cases. Some results on simulated data are computed and an application on temperate glaciers velocity estimation is processed. Finally, some additional improvements are performed: an adaptative sliding windows is set and a basic Bayes inference for flow model constraint is added.