Abstract:

On account of the scattering properties and target structure feature maintaining in polarimetric synthetic aperture radar (PolSAR) images after speckle suppressing, an effective algorithm for speckle reduction based on multi-stage classification is developed. First, the fast alternative to H/α is used in PolSAR image decomposition and the total back-scattering power is combined for initial pixel classification. Then minimum distance measure and hierarchical cluster method are used in the second stage classification. Finally, all pixels in the image are divided into three classes which are bright point or curve-linear targets, dark curve-linear targets, and other targets. The second and third kinds of targets are filtered by the linear minimize mean square filter (MMSE). Experimental results with AIRSAR data show that the new algorithm is more effective than Scattering-Model-Based speckle filter developed by Lee not only in speckle reduction but also in polarimetric properties and structure feature preservation.