Multi-look polar decomposition of polarimetric SAR images

This paper focuses an the multi-look polar decomposition of SAR images. Standard polar decomposition is generally used in single look polarimetry to decompose a bi-static or mono-static polarimetric scattering matrix into a product of an Hermitian matrix (boost) and a unitary matrix (rotation). An extension of this use in the framework of multi-looked polarimetry is proposed here. This new approach consists in decomposing the scattering matrix into boost and rotation components before vectorisation, then in averaging to generate boost and rotation coherency matrices separately, with new inferred parameters; the boost and rotation entropies, and concurrent dominant scattering mechanisms (alpha boost and alpha rotation). Tins multi-look extension of polar decomposition may allow for the definition of a new classification strategy for remote sensing data.