Wavelet-based Algorithm for Attenuation of Spatially Correlated Noise

摘要

This paper presents a wavelet-based algorithm to attenuate spatially correlated noise represented by a fractional Brownian motion. It generalizes the usual independence assumption by making the spatial relationship depend not only on the variance, but also on a long-memory parameter associated with the decay of autocorrelations. Wiener filtering in the wavelet domain obtains estimates of gray levels of the original signal. This provides a simple, fast and feasible solution for a denoising problem involving uncorrelated or spatially correlated noise.