Estimation of salt-pepper noise in images in wavelet domain

A novel approach was proposed for estimating the density of salt-pepper noise in images using wavelet transform. On the basis of the fact that the wavelet coefficients of all natural images conform to stable and close distribution, as well as such distribution of the noisy image may be influenced by the noise, the proposed algorithm exhibits how the wavelet coefficients magnitude histogram of the noisy image deviates from that of original image along with the density of the salt-pepper noise in quantitative form, and indicates that the degree of such deviation is nearly determined by the noise density, i.e., the change relation is robust to image traits. The proposed algorithm thus takes advantage of this relation to make estimation. Compared with those of existing methods, simulation results show that the proposed approach has more exact estimation value and less deviation.