Microseismic data denoising in the sychrosqueezed domain by integrating the wavelet coefficient thresholding and pixel connectivity

SUMMARY Microseismic monitoring is crucial for risk assessment in mining, fracturing and excavation. In practice, microseismic records are often contaminated by undesired noise, which is an obstacle to high-precision seismic locating and imaging. In this study, we develop a new denoising method to improve the signal-to-noise ratio (SNR) of seismic signals by combining wavelet coefficient thresholding and pixel connectivity thresholding. First, the pure background noise range in the seismic record is estimated using the ratio of variance (ROV) method. Then, the synchrosqueezed continuous wavelet transform (SS-CWT) is used to project the seismic records onto the time–frequency plane. After that, the wavelet coefficient threshold for each frequency is computed based on the empirical cumulative distribution function (ECDF) of the coefficients of the pure background noise. Next, hard thresholding is conducted to process the wavelet coefficients in the time–frequency domain. Finally, an image processing approach called pixel connectivity thresholding is introduced to further suppress isolated noise on the time–frequency plane. The wavelet coefficient threshold obtained by using pure background noise data is theoretically more accurate than that obtained by using the whole seismic record, because of the discrepancy in the power spectrum between seismic waves and background noise. After hard thresholding, the wavelet coefficients of residual noise exhibit isolated and lower pixel connectivity in the time–frequency plane, compared with those of seismic signals. Thus, pixel connectivity thresholding is utilized to deal with the residual noise and further improve the SNR of seismic records. The proposed new denoising method is tested by synthetic and real seismic data, and the results suggest its effectiveness and robustness when dealing with noisy data from different acquisition environments and sampling rates. The current study provides a useful tool for microseismic data processing.