ABSTRACT
In modern remote sensing, the use of synthetic aperture radar (SAR) represents an important source of information for Earth observation. Recent improvements have enabled modern satellite SAR missions, such as COSMO-SkyMed (CSK), TerraSAR-X, RADARSAT-2, to acquire high-resolution (HR) data (up to metric resolution) with a very short revisit time (e.g., 12 h for CSK). In addition, SAR is robust with respect to lack of illumination and atmospheric conditions. Together, these factors explain the rapidly growing interest in SAR imagery for various applications, such as ¤ood/ †re monitoring, urban mapping, and epidemiological surveillance. HR imagery allows to appreciate various ground materials resulting in highly mixed distributions. The resulting spatial heterogeneity is a critical problem in applications to image classi†cation (estimation of class-conditional statistics) or †ltering (estimation of local statistics, e.g., in moving-window approaches) [1]. Analysis and modeling of heterogenous HR SAR data pose a dif†cult statistical problem, which, to the best of our knowledge, has not been suf†ciently addressed so far.
