Speckle filtering of stationary and nonstationary scene signals in SAR imagery

Speckle filtering of nonstationary scenes is studied in the context of estimation theory. It is shown that the problem can be solved separately for stationary in increments scenes, and scenes which are not locally stationary. The speckle multiplicative noise model, the Kuan´s nonstationary mean nonstationary variance (NMNV) scene model, and the speckle-scene product model are analysed, and their use for an accurate reconstruction of the underlying scene signal is discussed. A protocol is introduced for speckle filtering of stationary and nonstationary scenes. It is shown that the use of multi-resolution algorithms is crucial for a safe reconstruction of the underlying nonstationary radar cross section of the illuminated scene