Abstract: For the inherent speckle noise in medical ultrasound images, this paper presents a denoising algorithm based on wavelet packet transform and nonhomomorphic filtering. In order to improve the capability of the separation of signal and noise, local correlation of wavelet packet decomposition coefficients is analysed in this paper. Then the normalized auto-covariance function of these above coefficients is computed. Thus the independent local observed value of each coefficient is achieved. Last a minimum mean square error estimation, which is combined with the local observed values and wavelet packet coefficients, is applied to get the nonhomomorphic filtering to denoise in wavelet packet domain. The experimental results of synthetic speckle and real ultrasound images show that the proposed algorithm outperforms several common medical ultrasound image denoising methods in terms of speckle reduction and edge preservation indices.