In order to preserve the textural feature affected by multiplicative speckle in high resolution synthetic aperture radar (SAR) images,a despeckling metod,based on contourlet domain adaptive BivaShrink denoising model,is proposed to suppress the speckle in SAR images. By calculating the variance homogeneous measurement and local energy ratio,the shape and size of variance estimating windows are determined adaptively,and then the variance of original images can be estimated optimally. Expriments show that the proposed algorithm performs better on not only speckle reduction but also preservation of structural detail information than wavelet BivaShrink algorithm and contourlet threshold algorithm. Visual effect and experimental numerical index both are improved apparently.