Non-Homogeneous Wavelet Denoising
Sinisa Pajevic, Gustavo Kunde Rohde, Peter J. Basser, Akram Aldroubi · 2004
Introduction Wavelet denoising has been widely used in signal processing to remove noise while preserving edges and spikes in the data, with minimal smoothing. A standard approach is wavelet shrinkage (WS) of the wavelet coefficients, dj, j = 1,...,n, [1,2] using a threshold t =σ 2 log n , a decision theoretic result based on risk minimization. Another approach is to treat thresholding as a hypothesis testing problem [3,4], with null hypothesis H0: d j t = 0. Both approaches address the case of signals corrupted by additive and spatially homogeneous noise. Often, noise properties are