Seismic random noise attenuation using modified wavelet thresholding

Abstract

<p>In seismic exploration, random noise deteriorates the quality of acquired data. This study analyzed existing denoising methods used in seismic exploration from the perspective of random noise. Wavelet thresholding offers a new approach to reducing random noise in simulation results, synthetic data, and real data. A modified wavelet threshold function was developed by considering the merits and demerits of conventional soft and hard thresholding schemes. A MATLAB (matrix laboratory) simulation model was used to compare the signal-to-noise ratios (SNRs) and mean square errors (MSEs) of the soft, hard, and modified threshold functions. The results demonstrated that the modified threshold function can avoid the pseudo-Gibbs phenomenon and produce a higher SNR than the soft and hard threshold functions. A seismic convolution model was built using seismic wavelets to verify the effectiveness of different denoising methods. The model was used to demonstrate that the modified thresholding scheme can effectively reduce random noise in seismic data and retain the desired signal. The application of the proposed tool to a real raw seismogram recorded during a land seismic exploration experiment located in north China clearly demonstrated its efficiency for random noise attenuation.</p>

Using geometric mode decomposition for the background noise suppression on microseismic data

“…Forghani et al., 2012; Sabbione et al., 2013, 2015; Abbasi & Ismail, 2021; Abbasi & Jaiswal, 2013), wavelet transform (e.g. Goudarzi & Riahi, 2012; Li et al., 2018; Mousavi et al., 2016; Yoon & Vaidyanathan, 2004; Zhang et al., 2016), S ‐transform (e.g. Akram, 2018, 2020; Parolai, 2009), curvelet transform (e.g.…”

An attention‐guided convolution neural network for denoising of distributed acoustic sensing–vertical seismic profile data

“…Aiming at the suppression of noise and the retention of effective signals, researchers have constantly put forward new approaches and made some achievements (Gholami, 2014; Wang et al., 2020; Zhang et al., 2017; Zheng et al., 2015). Band‐pass filtering cannot remove the noise close to or coincident with the effective signal frequency; the effect of noise suppression of F–X deconvolution is poor (Xu et al., 2016); the time‐frequency analysis approach has not been widely used in seismic exploration signal processing.…”