Prior scene knowledge for the Bayesian restoration of mono-and multi-channel SAR images

Ideally, using SAR data in combination with optical data or to invert of a physical backscattering model, prior scene knowledge is introduced in adaptive speckle filters to aim at restoration of radar reflectivity i.e. of the physical quantity, proportional to the backscattering coefficient, that is measured by a SAR instrument. Introduction of A Priori knowledge or A Priori guess implies generally the use of Bayesian methods in the processing of SAR images. In this paper, we analyse how prior knowledge, or prior guess, of the first order and of second order statistics of the imaged scene has been gradually introduced in the development of adaptive speckle filters. It is shown how these scene statistical models are used, in particular in a Bayesian Maximum A Posteriori (MAP) inference process. These Bayesian filters, that present the structure of control systems, are analysed in terms of stability and commandability.