To remove noises without blurring the image edge, an adaptive method for image denoising with nondecimated wavelet and edge preservation is proposed in this paper. Based on the interseale dependency of wavelet coefficients,we make a difference between the wavelet coefficients related to noise,the coeffidents associated to homogeneous regions and the coefficients noise related. At each resolution coefficients and the noise-related coeffdents are modeled by different statistical models. The wavelt coefficients related to homogeneous regions are processed by an adaptive threshold,which is derived from Bayesian estimation. Experimental results show that this method often obtains clearer edges and better denoising perform ance in comparison with common methods.