A novel mean filter based on the partial distribution for SAR images speckle reduction

Traditional filtering methods can´t work well in Synthetic Aperture Radar (SAR) image speckle reduction due to the existence of the speckle noise. The MMSE-based filter, such as Frost filter, Lee filter and Kuan filter, assumes that noise image and observed image obey to the Gaussian distribution. In fact, the assumption of Rayleigh, Gamma, Weibull distribution is more consistent with SAR images. Those reasoned distribution of assumption are all partial distribution. In this paper, a novel mean filter to reduce SAR images speckle is designed by analyzing the partial distribution statistical characteristic of SAR images and existing spatial filtering methods. The novel method is evaluated and compared with well-accepted methods through homogeneous and heterogeneous SAR images. The evaluated results display Its well efficiency and adaptability.