Synthetic Aperture Radar (SAR) image despeckling is an important problem in the SAR applications. The demand for the speckle reduction of SAR images is to smooth the speckle noise while preserving the structure information of the original images. Because contourlet is a new and more effective signal representation tool than wavelet in many image applications, we propose an improved despeckling method based on the hidden Markov tree (HMT) model in contourlet domain. In the new method, a parameter named inter-direction variation coefficient which takes advantages of both inter-direction dependency of the contourlet coefficients and measurement of the scene heterogeneity is developed to upgrade the performance. Experiments on the real SAR images show that the proposed method achieves better performance in contrary to other despeckling methods.