A noise reduction approach based on stein's unbiased risk estimate

Abstract

This paper proposes a new wavelet-based shrinkage function for 1D signal noise reduction. This shrinkage function adopts the intrascale correlations between wavelet coefficients and exploits Stein's unbiased risk estimator to achieve the optimal parameter. Unlike the methods based upon Bayes estimators, the proposed method does not use any prior hypotheses on wavelet coefficients. Experiments performed on simulated signals clearly indicate that our method outperforms conventional noise reduction methods in the sense of the signal-to-noise ratio.