Denoising of low SNR signals using composite wavelet shrinkage
R.S.C. Wong, V.K. Bhargava · 2002
This paper introduces a composite method for the calculation of thresholds the for denoising of signals, along with the concept of clipping. This composite method is based on the universal and level dependant methods. The proposed method uses a modified level dependant threshold for levels with more than a specified number of coefficients, s, and a modified universal threshold, instead of 0, for levels with fewer than s coefficients. If the modified level dependant threshold exceeds some clipping value, it is replaced by the modified universal threshold. The results show that for low SNR signals corrupted by white noise, the proposed thresholding technique outperforms both the level dependant and universal thresholds for a variety of signals.