In this chapter, we fi.rst address the general problem of modeling the statistics of single-channel synthetic aperture radar (SAR) amplitude images and, speci.cally, high resolution (HR) SAR. Given the variety of the existing approaches, we extend and enhance the discionary-based stochastic expecation-maximization (DSEM) technique proposed previously for coarser-resolution SAR. We expect DSEM to be an appropriate tool for this modeling problem, since it is an intrinsically flexible method, modeling SAR statistics as resulting from mixing several populations, and it is not constrained to a speci.c choice of a given parametric model allowing to benefi.t from many of them (dictionary approach). Thus, we extend the earlier DSEM approach to HR satellite SAR imagery and enhance it by introducing a novel procedure for estimating the number of mixture components, which enables to appreciably reduce its computational complexity, resulting in an Enhanced DSEM algorithm (EDSEM). Building on the proposed method for single channel SAR pdf estimation we proceed to multi-channel joint pdf estimation and classi.cation. Contemporary satellite SAR missions are capable of registering polarimetric (PolSAR) imagery, which provides a more complete description of landcover scattering behavior than single-channel SAR data. The potential for improved classi.fication accuracy with data in several polarizations, compared to single-channel data, explains the special interest to polarimetric SAR image classi.cation. Furthermore, several current satellite SAR systems, e.g., TerraSAR-X, COSMO-SkyMed, RADARSAT-2, support, at least, dual-pol acquisition modes. In this chapter we investigate the quad polarization (quad-pol), dual polarization (dual-pol), as well as single polarization (single-pol) SAR imagery scenarios.