Context

Full polarimetric data provide an unique possibility to separate scattering contributions of different nature, which can be associated to certain elementary scattering mechanisms. Several decomposition techniques have been proposed for this purpose. Two main classes of decompositions can be identified Fig. 4.1.

Figure 4.1: Overview of Decomposition Theorems (by Eric Pottier)

One which deals with decomposition of the complex voltage reflection (Sinclair) matrix, another which deals with decomposition of power reflection (Müller/Kennaugh and covariance type) matrices. Decompositions based on the complex voltage reflection matrix are for example, the Pauli decomposition and the sphere/diplane/helix decomposition which shall be discussed in the following. The decompositions based on the decomposition of power reflection can be divided into approaches based on the Kennaugh matrix (e.g. Huynen decomposition) and approaches based on the covariance [C] or coherency matrix [T]. Traditionally, the approaches based on the Müller/Kennaugh matrix preferred in optics, while for radar remote sensing the covariance [C] or coherency matrix [T] based approaches are usually preferred.

The classical and first considerations about radar target decomposition techniques were published in [Huynen70]. The basic idea behind this approach is to decompose the averaged Müller/Kennaugh matrix into a sum of a deterministic single target component matrix "single" and a distributed residue component matrix ("N-target, where N refers to, but should not too literally regarded as, noise) which is related to non-symmetric scattering contributions. The approach is not capable of yielding uniquely the dominant scattering process in different polarization bases [Cloude92]. Due to this fact, and because of their more compact form, approaches based on the coherence and covariance matrices have been preferred in recent years. Therefore, in this treatment we will only focus on approaches based on the Sinclair ([S]) - matrix. For the sake of completeness, we should mention, that the display of decomposition techniques in this treatment is neither complete nor exhaustive. We merely put together the most commonly used techniques. There are also some model based techniques existent (e.g. A. Freeman and J. van Zyl) but these are beyond the scope of this treatment.