A Laplace and generalized Gaussian mixture distribution-based method is proposed to explore the nonsubsampled Contourlet transform (NSCT) coefficients. The investigating result reveals that the NSCT,as a shift-invariant contourlet transform,obtains redundant coefficients in each scale and each direction,and its coefficients differ from those of Contourlet transform in aspects of general Gaussian distribution. The regularized parameters should be introduced to generalized Gaussian distribution model to enhance the coefficients distribution. The medical image denoising experimental results with comparison to similar Contourlet-based methods indicate the proposed modeling method improves the accuracy of noise estimation,increases the peak signal-noise ratio,and achieves better visual quality.