<FONT face=Verdana>Synthetic aperture radar (SAR) images are inherently degraded by speckle noise

that severely affects the image qualities and makes the automatic interpretation of the image data very difficult. Combine the SAR image statistical property with the favorable capability of nonsubsampled contourlet transform on describing SAR images detail information

a novel SAR image despeckling method is presented. We estimate the speckle variance in each high-frequency directional subband. The local directional statistical information and the estimated speckle variance is used to classify the nonsubsampled contourlet transform coefficients as signal and noise coefficients. The SAR image despeckling is implemented by retaining signal coefficients and restraining noise coefficients. Experimental results demonstrate that the proposed despeckling method can preserve detail information effectively and reduce the speckle noise at the same time.