Signal Denoising Using Simplified SURE-Based Linear Expansion of Thresholds in the Wavelet Domain

The purpose of signal denoising is to reduce the noise level, while preserving the signal features. Various denoising approaches exist, in which wavelet-based pointwise thresholding approaches are extensively adopted in many application fields. For signal denoising based on wavelet thresholding, there are two crucial aspects, namely, the use of a proper thresholding function and the estimate of the noise standard deviation. Both greatly influence the quality of the denoised signal. In this paper, a simple wavelet-based denoising approach is proposed, which uses the modified linear expansion of thresholds based on Stein’ s unbiased risk estimate (SURE) and the noise standard deviation’s estimation depending on the number of vanishing moments of the wavelet transform. Experimental results demonstrate that higher signal-to-noise ratio (SNR) can be obtained by this new approach as compared to other thresholding approaches.