In order to improve some shortcoming in the soft threshold and hard threshold and low threshold functions have improved of wavelet threshold denoising, the authors propose a new hierarchical denoising threshold. Firstly, the method decomposes the noise image. Thus it is concluded that wavelet coefficients. Secondly, the wavelet coefficient by using an improved threshold function was carried out on the high-frequency part of the threshold processing. Finally, according to the estimation in the wavelet base conditions, and to rebuild image, this method gets the denoised image. The method has excellent mathematical characteristics, through the medical image simulation results indicate that the de-noising method effect both in the visual effects. Or in the MSE and SNR performance analysis is better than above mentioned denoising methods. So the algorithm in solving the problem of actual denoising is worthy of popularization and application.