My PhD
My Ph.D. deals with the classification of very high resolution remote sensing images.
The considered method combines a statistical modeling of the input data (SAR, optical) - which uses a dictionary-based mixture probability density function (PDF) estimation approach - with a Markov random field (MRF) model, thus allowing to take into account the contextual information. The algorithm robustness has been increased thanks to the introduction of textural features extracted from the original SAR images.
The method has been extended to multiresolution, multiband and/or multisensor acquisitions by resorting to the use of a hierarchical Markov random field model based on a quad-tree structure, where each tree scale corresponds to the different input image resolutions. To obtain the classification map, we integrate an exact estimator of the marginal posterior mode.
This Ph.D. is in collaboration with the University of Genoa (Italy).