ARTICLE

Desert seismic data denoising and effective signal recovery by using improved shearlet transform based on the deep-learning coefficient selection

XINTONG DONG1 YUE LI2* BAOJUN YANG3
Show Less
1 Jilin University, College of Instrumentation and Electrical Engineering, Jilin, P.R. China.,
2 Jilin University, College of Communication Engineering, Jilin, P.R. China.,
3 Jilin University, College of Geo-exploration Science and Technology, Jilin, P.R. China.,
JSE 2021, 30(5), 455–479;
Submitted: 18 August 2020 | Accepted: 10 June 2021 | Published: 1 October 2021
© 2021 by the Authors. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution -Noncommercial 4.0 International License (CC-by the license) ( https://creativecommons.org/licenses/by-nc/4.0/ )
Abstract

Dong, X., Li, Y. and Yang, B., 2021. Desert seismic data denoising and effective signal recovery by using improved shearlet transform based on the deep-learning coefficient selection. Journal of Seismic Exploration, 30: 455-479. Contamination of seismic data by background noise causes difficulties for imaging, reservoir fluid prediction, and stratigraphic interpretation. Desert seismic data poses a particular problem mainly due to two reasons: (1) low signal-to-noise ratio (SNR); (2) serious frequency spectrum overlapping between the effective signals and low-frequency noise (mainly including random noise and surface waves). Therefore, when apply sparse-transform-based methods to denoise desert seismic data, conventional threshold functions fail to distinguish the effective signal coefficients and low-frequency noise coefficients, which is likely to result in residual noise and signal leakage. To solve this problem, we utilize the convolutional neural network (CNN) to act as a threshold function, thereby establishing an optimal non-linear relationship between noisy coefficients and effective signal coefficients. In addition, in order to achieve multi-scale and multi-direction accurate noise suppression, we construct a corresponding training dataset for each sub-band, so as to obtain a CNN-based coefficient selection model suitable for this sub-band. In this paper, we take shearlet transform as an example to verify the effectiveness of the proposed CNN-based threshold function. Synthetic and real examples demonstrate that our method can effectively suppress the desert low-frequency noise and completely recover the effective signals reflected by layers.

Keywords
desert seismic data
shearlet transform
noise suppression
convolutional neural network
low signal-to-noise ratio
spectrum overlapping.
References
  1. Akram, J., 2018. An application of waveform denoising for microseismic data usingpolarization-linearity and time-frequency thresholding. Geophys. Prosp., 66:872-893.
  2. Bekara, M. and van der Baan, M., 2007. Local singular value decomposition for signalenhancement of seismic data. Geophysics, 72(2): V59-V65.
  3. Bekara, M. and van der Baan M., 2009. Random and coherent noise attenuation byempirical mode decomposition. Geophysics, 74(5): V89-V98.
  4. Beckouche, S. and Ma, J.W., 2014. Simultaneous dictionary learning and denoising forseismic data. Geophysics, 79(3): A27-A31.
  5. Bonar, D. and Sacchi, M., 2012. Denoising seismic data using the nonlocal meansalgorithm. Geophysics, 77(1): A5-A8.
  6. Chen, Y.K., Zhang, M., Bai, M. and Chen, W., 2019. Improving the signal-to-noise ratioof seismological datasets by unsupervised machine learning. Seismol. Res. Lett.,90: 1552-1564.
  7. Dong, X.T., Li, Y. and Yang, B.J., 2019a. Desert low-frequency noise suppression byusing adaptive DnCNNs based on the determination of high-order statistic.Geophys. J. Internat., 219: 1281-1299.
  8. Dong, X.T., Jiang, H., Zheng, S., Li, Y. and Yang, B.J., 2019b. Signal-to-noise ratioenhancement for 3C downhole microseismic data based on the 3D shearlet transformand improved back-propagation neural networks, Geophysics 84(4),V245-V254.
  9. Dong X.T., Zhong T. and Li Y., 2020a. New suppression technology for low-frequencynoise in desert region: the improved robust principle component analysis based onprediction of neural network. IEEE Transact. Geosci. Remote Sens., in Press.
  10. Dong, X.T., Zhong, T. and Li, Y., 2020b. A deep-learning-based denoising method formultiarea surface seismic data. IEEE Geosci. Remote Sens. Lett., in Press.
  11. Duncan, G. and Beresford, G., 1995. Median filter behaviour with seismic data.Geophys. Prosp., 43: 329-345.
  12. Fomel, S. and Liu, Y., 2010. Seislet transform and seislet frame. Geophysics, 75(3):V25-V38.
  13. Gholami, A. and Zand, T., 2017. Fast L1-regularized radon transforms for seismic dataprocessing. Digit. Sign. Process., 71: 83-94.
  14. Goudarzi, A. and Riahi, M.A., 2012. Seismic coherent and random noise attenuationusing the undecimated discrete wavelet transform method with WDGA technique. J.Geophys. Engineer., 9: 619-631.
  15. Guo, K. and Labate, D., 2007. Optimally sparse multidimensional representation usingshearlets. SIAM J. Mathemat. Anal., 39: 298-318.
  16. Giiltinay, N., 2017. Signal leakage in f -x deconvolution algorithms. Geophysics, 82(5):W31-W45.
  17. Han, J.J. and van der Baan, M., 2013. Empirical mode decomposition for seismictime-frequency analysis. Geophysics, 78(2). 09-019.
  18. Harris, P.E. and White, R.E., 1997. Improving the performance of f-x prediction filteringat low signal-to-noise ratios. Geophys. Prosp., 45: 269-302.
  19. Jin, K.H., McCann, M.T., Froustey, E. and Unser, M., 2017. Deep convolutional neuralnetwork for inverse problems in imaging. IEEE Transact. Image Process., 26:4509-4522.
  20. Kulesh, M., Diallo, M., Holschneider, M., Kurennaya, K., Kruger, F., Ohrnberger, M.and Scherbaum, F., 2007. Polarization analysis in the wavelet domain based onadaptive covariance method. Geophys. J. Internat., 170: 667-678.
  21. Kutyniok, G. and Lim, W., 2011. Compactly supported shearlets are optimally sparse. J.Approxim. Theory, 163: 1564-1589.
  22. Li, F., Zhang, B., Verma, S. and Marfurt, K.J., 2017. Seismic signal denoising usingthresholded variational mode decomposition. Explor. Geophys., 49: 450-461.
  23. Li, G.H., Li, Y. and Yang, B.J., 2017. Seismic exploration random noise on land:
  24. Modeling and application to noise suppression. IEEE Transact. Geosci. RemoteSens., 55: 4668-4681.
  25. Liang, X.Q., Li, Y. and Zhang, C., 2018. Noise suppression for microseismic data bynon-subsampled shearlet transform based on singular value decomposition.Geophys. Prosp., 66: 894-903.
  26. Lim, W., 2010. The discrete shearlet transform: A new directional transform andcompactly supported shearlet frames. IEEE Transact. Image Process., 19,1166-1180.
  27. Loffe, S. and Szegedy, C., 2015. Batch normalization: Accelerating deep networktraining by reducing internal covariate shift. Proc., 32nd Internat. Conf. Mach.Learning, 448-456.
  28. Ma, H.T, Yao, H.Y., Li, Y. and Wang, H.Z., 2020a. Deep residual encoder-decodernetworks for desert seismic noise suppression. IEEE Geosci. Remote Sens. Lett.,17: 529-533.
  29. Ma, H.T, Yan, J. and Li, Y., 2020b. Low-frequency noise suppression of desert seismicdata based on variational mode decomposition and low-rank component extraction.IEEE Geosci. Remote Sens. Lett., 17: 337-341.
  30. Mou, L.C., Ghamisi, P. and Zhu, X.X., 2018. Unsupervised spectral-spatial featurelearning via deep residual conv-deconv network for hyperspectral imageclassification. IEEE Geosci. Remote Sens. Lett., 56: 391-406.
  31. Nazari Siahsar, M.A., Gholtashi, S., Abolghasemi, V. and Chen, Y., 2017. Simultaneousdenoising and interpolation of 2D seismic data using data-driven non-negativedictionary learning. Sign. Process., 141: 309-321.
  32. Neelamani, R., Baumstein, A.I., Gillard, D.G., Hadidi, M.T. and Soroka, W.L., 2008.
  33. Coherent and random noise attenuation using the curvelet transform. The LeadingEdge, 27: 240-248.
  34. Remez, T., Litany, O., Giryes, R. and Bronstein, A.M., 2018. Class-aware fullyconvolutional Gaussian and Poisson denoising. IEEE Transact. Geosci. RemoteSens., 27: 5707-5722.
  35. Stein, R.A. and Bartley, N.R., 1983. Continuously time-variable recursive digitalband-pass filters for seismic signal processing. Geophysics 48: 702-712.
  36. Tang, N., Zhao, X., Li, Y. and Zhu, D., 2018. Adaptive threshold shearlet transform forsurface microseismic data denoising. J. Appl. Geophys., 153: 64-74.
  37. Tian, Y.N., Li, Y. and Yang, B.J., 2014. Variable-eccentricity hyperbolic-trace TFPF forseismic random noise attenuation. IEEE Transact. Geosci. Remote Sens., 52:6449-6458.
  38. Tirer, T. and Giryes, R., 2019. Super-resolution via image-adapted denoising CNNs:
  39. Incorporating external and internal learning. IEEE Signal Processing Letters 26(7),1080-1084.
  40. Wang, F. and Chen, S., 2019. Residual learning of deep convolutional neural network forseismic random noise attenuation. IEEE Geosci. Remote Sens. Lett., 16: 1314-1318.
  41. Wang, Y., 1999. Random noise attenuation using forward-backward linear prediction. J.Seismic Explor., 8: 133-142.
  42. Wang, Y., Liu X., Gao F. and Rao Y., 2020. Robust vector median filtering with astructure-adaptive implementation. Geophysics, 85(5): V407-V414.
  43. Wang, Y., Lu, W., Wang, B. and Liu, L., 2016. A robust polynomial principalcomponent analysis for seismic noise attenuation. J. Geophys. Engineer., 13:1002-1009.
  44. Wu, S., Li, G.Q., Deng, L., Liu, L., Wu, D., Xie, Y. and Shi, L.P., 2019. L1-norm batchnormalization for efficient training of deep neural networks. IEEE Transact. NeuralNetw. Learning Syst., 30: 2043-2051.
  45. Wu, Y. and McMechan, G.A., 2019. Parametric convolutional neural network-domainfull-waveform inversion. Geophysics, 84(6): R881-R896.
  46. Xiong, M.J., Li, Y. and Wu, N., 2014. Random-noise attenuation for seismic data bylocal parallel radial-trace TFPF. IEEE Transact Geosci. Remote Sens., 52:4025-4031.
  47. Yang, W., Feng, J., Yang, J., Zhao, F., Liu, J., Guo, Z. and Yan, S., 2017. Deep edgeguided recurrent residual learning for image super-resolution. IEEE Transact. ImageProcess., 26: 5895-5907.
  48. Yang Y., Liu C. and Langston C.A., 2020. Processing seismic ambient noise data withthe continuous wavelet transform to obtain reliable empirical Green's functions.Geophys. J. Internat., 222: 1224-1235.
  49. Yuan, S.Y., Liu, J.W., Wang, S.X., Wang, T.Y. and Shi, P.D., 2018. Seismic waveformclassification and first-break picking using convolution neural networks. IEEEGeosci. Remote Sens. Lett., 15: 272-276.
  50. Zhang, K., Zuo, W.M., Chen, Y.J., Meng, D.Y. and Zhang, L., 2017. Beyond a Gaussiandenoiser: Residual learning of deep CNN for image denoising. IEEE Transact.Image Process., 26: 3142-3155.
  51. Zhong, T., Li, Y., Wu, N., Nie, P.F. and Yang, B.J., 2015a. Statistical analysis ofbackground noise in seismic prospecting. Geophys. Prosp., 63: 1161-1174.
  52. Zhong, T., Li, Y., Wu, N., Nie, P.F. and Yang, B.J., 2015b. A study on the stationarityand Gaussianity of the background noise in land-seismic prospecting. Geophysics,80(4): V67-V82.
Share
Back to top
Journal of Seismic Exploration, Electronic ISSN: 0963-0651 Print ISSN: 0963-0651, Published by AccScience Publishing