Wavelet based seismic signal de-noising using Shannon and Tsallis entropy

Seismograms are the vital sources of information in seismic engineering. But, these records are always contaminated with noise which has to be removed before using them in seismic applications. Recently, wavelet based techniques proved to be very effective in de-noising by achieving high SNR. However, selection of the correct threshold plays a crucial role in deciding the SNR value. It is strange that only very few thresholders exist in seismic and non-seismic studies. In this paper, we have proposed a set of novel entropy based thresholders through 2 experiments. In experiment 1, we have proposed a Shannon entropy based algorithm which has produced 11.205 SNR. In experiment 2, we used Tsallis entropy which has moderately improved the result by providing 12.23 SNR. Existing thresholders like visu and normal shrink have managed to produce 10.19 and 10.07 SNR respectively. Through our experiments, we observed that for low frequency problems (σ=0.27σ=0.27), the performance of both entropies matched appreciably. However, for high frequency (σ=2.7σ=2.7) Tsallis produced slightly better SNR and is more feasible in detecting the occurrence of P and S waves by smoothing the accelerograms.