基于平移不变小波的原子力图像去噪算法研究

In this paper, we analyze the noise characteristics of images obtained from the AFM, and it shows the noise in each image to have a non-uniform Gaussian distribution. According to the difference of image noise propagation in each subband, a novel multiscale adaptive thresholding based on translation invariant(TI) wavelet transform has been proposed. And we also give the general framework for image denosing based on TI wavelet transform. Experiments show that the proposed method can choose the threshold adaptively. Its efficiency is shown by comparison with the classical Bayesian and the Visulthrink based methods for Lena and AFM image denoising.