Abstract: A fast algorithm based on the automatic censoring is proposed to filter the speckle in POLSAR image. Firstly, a rectangle widow is sliding through each pixel of POLSAR imagery. For each current pixel, a global threshold is used to determine whether the current pixel is a point target. If not, then the homogeneous pixels of the same main scattering mechanism with this current pixel are censored automatically in the current sliding-window. Finally, the locally linear minimum mean-squared error (LLMMSE) estimator is used to filter this pixel with the selected pixels. The selection of homogeneous pixels based on the ML texture reduces the effect of speckle; the selection of the same main scattering mechanism pixels based on scattering similarity preserves the main scattering mechanism. The experiment results of a typically actual POLSAR scene and theoretical analysis show that the proposed algorithm has the well theoretical and actual performance in speckle filtering, texture preservation and target main scattering preservation.