Classified pixel-based windowing algorithm for polarimetric SAR speckle filtering

Two main purposes of synthetic aperture radar (SAR) speckle filtering are the preservation of the edge sharpness and the suppression of the speckle noise. However, most of the existing filtering techniques suppress the speckle noise while smearing edges. This problem is partly solved by using nonsquare windows, but it can also generate large errors in many applications. A new windowing algorithm that can overcome the defects of the nonsquare window, by discriminating the edges more precisely with a preliminary terrain classification result, is proposed.