A simple method inspired by empirical mode decomposition for denoising seismic data

Abstract

We developed a new and simple method for denoising seismic data, which was inspired by data-driven empirical mode decomposition (EMD) algorithms. The method, which can be applied either as a trace-by-trace process or in the f-x domain, replaces the use of the cubic interpolation scheme, which is required to calculate the mean envelopes of the signal and the residues, by window averaging. The resulting strategy is not viewed as an EMD per se, but a userfriendly version of EMD-based algorithms that permits us to attain, in a fraction of the time, the same level of noise cancellation as standard EMD implementations. Furthermore, the proposed method requires less user intervention and easily processes millions of traces in minutes rather than in hours as required by conventional EMD-based techniques on a standard PC. We compared the performance of the new method against standard EMD methods in terms of computational cost and signal preservation and applied them to denoise synthetic and field (microseismic and poststack) data containing random, erratic, and coherent noise. The corresponding f-x EMDs implementations for lateral continuity enhancement were analyzed and compared against the classical f-x deconvolution to test the method.

Signal Denoising Method Combined With Variational Mode Decomposition, Machine Learning Online Optimization and the Interval Thresholding Technique

“…Predict the distribution of F(p * ), p * =α * . For any p * in Ω, which is the space of all possible parameter values, we can predict the distribution of F(p * ) with (12) and 13based on the model built in step 1.…”

“…The time complexity of VMD is O (2Nlog 2 (2N) In order to verify the correctness of proposed algorithm complexity, a series of experiments are carried out. The test signal is Bumps with 5 dB of Gaussian white noise, where the length N ranges from 2 8 to 2 12 . The configuration of the computer is as follows: Intel (R) Pentium (R) G3260 @3.30GHz CPU and 4.00 GB RAM memory running windows 7.…”

“…EMD is a recursive decomposition model that can decompose a signal into a finite number of intrinsic mode functions (IMFs) with an iterative process called sifting [10], [11]. The decomposition is related to the characteristics of the signal itself without the requirement of any prior basis functions [12]. With these advantages, many EMD-based denoising methods have been proposed, such as EMD-HT, EMD-ST [13], and EMD-STRP [14].…”

Convolutional neural network‐based classification of microseismic events originating in a stimulated reservoir from distributed acoustic sensing data

“…Historically, microseismic event detection has been approached using techniques such as short‐term average to long‐term average (STA/LTA) (Withers et al ., 1998), cross‐correlation of waveforms (Gibbons & Ringdal, 2006) and power spectral density measurement (Vaezi & van der Baan, 2015). In regard to seismic data processing, techniques such as prediction filtering (Canales, 1984), composite property‐mapping signal enhancement (Cadzow, 1988), the curvelet transform (Candès et al ., 2006) and empirical mode decomposition (Gomez & Velis, 2016) have been utilized in order to improve signal‐to‐noise ratios.…”

A Nonparametric Method for Automatic Denoising of Microseismic Data

“…Hence, numerous methods have been presented for this in the past few decades that can be classified into four main categories: (1) infinite impulse response (IIR) filters, known as low-pass, high-pass, band-pass, and band-stop filter (e.g., [5]); (2) Fourier-based filters (e.g., [6][7][8]); (3) EMD-based (empirical mode decomposition) filters (e.g., [9][10][11][12][13][14]); (4) wavelet-based filters (e.g., [15][16][17]). …”