Classification of multi-sensor remote sensing images using an adaptive hierarchical Markovian model

In this paper, we propose a novel method for the classification of the multi-sensor remote sensing imagery, which represents a vital and fairly unexplored classification problem. The proposed classifier is based on an explicit hierarchical graph-based model sufficiently flexible to deal with multi-source coregistered datasets at each level of the graph. The suggested supervised method relies on a two-step technique. In the first step, a joint statistical model is developed for the input images that consists of the finite mixtures of automatically chosen parametric families for single images, and multivariate copulas to model joint class-conditional statistics at each resolution. As a second step, we plug the estimated joint probability density functions into a hierarchical Markovian model based on a quad-tree structure. Multi-scale features correspond to different resolution images or are extracted by discrete wavelet transforms. To obtain the classification map, we resort to an exact estimator of the marginal posterior mode.