Filtering of polarimetric synthetic aperture radar images: A sequential approach

This paper introduces a novel method for filtering polarimetric synthetic aperture radar (POLSAR) imagery data. Unlike most existing filters, the proposed one is freed from using searching windows but implements the filtering in a sequential fashion. Specifically, we propose to use the inverse Wishart distribution as the conjugate prior model for the coherency matrix (CM) data and show that the maximum a posteriori (MAP) solution leads to a very simple linear recursive formula for estimate update, through which issues such as feature preservation and bias correction can be also taken into account. In the experiment, the proposed method is tested on the POLSAR dataset acquired by the DLR (Germany Aerospace Center) F-SAR system. Results demonstrate that the filter shows inspiring performances in terms of both noise reduction and feature preservation.