ARTICLE

Seismic data denoising using double sparsity dictionary and alternating direction method of multipliers

LIANG ZHANG1,2 LIGUO HAN1 AO CHANG1 JINWEI FANG3 PAN ZHANG1 YONG HU1 ZHENGGUANG LIU2
Show Less
1 College of Geo-exploration Science and Technology, Jilin University, Changchun 130026, P.R. China.,
2 Institute of Geosciences and Info-Physics, Central South University, Changsha, Hunan 410083, P.R. China.,
3 CNPC Key Lab of Geophysical Exploration, China University of Petroleum-Beijing, Beijing 102249, P.R. China.,
JSE 2020, 29(1), 49–71;
Submitted: 3 August 2018 | Accepted: 11 October 2019 | Published: 1 February 2020
© 2020 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

Zhang, L., Han, L.G., Chang, A., Fang, J.W., Zhang, P., Hu, Y. and Liu, Z.G., 2020. Seismic data denoising using double sparsity dictionary and alternating direction method of multipliers. Journal of Seismic Exploration, 29: 49-71. Recently, the dictionary learning plays a more and more important role in seismic ata denoising. Compared with the fixed-basis transform (e.g., Fourier transform, wavelet ansform, curvelet transform, contourlet transform and shearlet transform), the denoising f dictionary learning is better because of adaptive sparse representation of seismic data. owever, dictionary learning often produces artifacts due to no prior-constraint structural nformation. In this paper, we propose a new denoising approach, which has double sparsity and combines the advantage of fixed-basis transform and dictionary learning. The whole work-flow of the new denoising approach is as follows. Firstly, we can obtain sparse coefficients of seismic data via shearlet transform. Secondly, sparse coefficients are divided into some suitable size blocks which are regarded as training sets. Thirdly, the alternating direction method of multipliers (ADMM) is used in sparse coding to update ictionary coefficients. Then, the data-driven tight frame (DDTF) is used in dictionary updating to update dictionary atoms. Again, the ADMM is used to resolve the convex optimization problem, and we reshape output blocks to obtain new sparse coefficients. Finally, the hard-thresholding and inverse shearlet transform are applied to new sparse oefficients to achieve denoising. The synthetic data and field data experiments show that e new denoising approach obtain better result than fixed-basis transform and dictionary earning. In conclusion, the new denoising approach can attenuate artifacts and improve e quality of seismic data denoising.

Keywords
seismic data denoising
shearlet transform
DDTF
ADMM
References
  1. Aharon, M., Elad, M. and Bruckstein, A., 2006. $ rm k $-SVD: An algorithm fordesigning overcomplete dictionaries for sparse representation. IEEE Transact.Sign. Process., 54: 4311-4322.
  2. Anvari, R., Siahsar, M.A.N., Gholtashi, S., Kahoo, A.R. and Mohammadi, ?? 2017.
  3. Seismic random noise attenuation using synchrosqueezed wavelet transform andlow-rank signal matrix approximation. IEEE Transact. Geosci. Remote Sens., 55:6574-6581.
  4. Bertsekas, D.P., 1999. Nonlinear Programming. Belmont: Athena Scientific, Boston, MA.
  5. Boyd, S., Parikh, N., Chu, E., Peleato, B. and Eckstein, J., 2011. Distributed optimizationand statistical learning via the alternating direction method of multipliers. Foundat.Trends Mach. Learn., 3: 1-122.
  6. Bunks, C., Saleck, F.M., Zaleski, S. and Chavent, G., 1995. Multiscale seismicwaveform inversion. Geophysics, 60: 1457-1473.
  7. Cai, J.F., Ji, H., Shen, Z. and Ye, G.B., 2014. Data-driven tight frame construction andimage denoising. Appl. Computat. Harm. Analys., 37: 89-105.
  8. Cao, J., Zhao, J. and Hu, Z., 2015. 3D seismic denoising based on a low-redundancycurvelet transform. J. Geophys. Engineer., 12: 566.
  9. Chen, Y., Fomel, S. and Hu, J., 2014. Iterative deblending of simultaneous-sourceseismic data using seislet-domain shaping regularization. Geophysics, 79(5):V179-V 189.
  10. Chen, Y., Zhang, L. and Mo, L., 2015. Seismic data interpolation using nonlinearshaping regularization. J. Seismic Explor., 24: 327-342.
  11. Chen, Y., Ma, J. and Fomel, S., 2016. Double-sparsity dictionary for seismic noiseattenuation. Geophysics, 81(2): V103-V116.
  12. Chen, Y., 2017. Fast dictionary learning for noise attenuation of multidimensionalseismic data. Geophys. J. Internat., 209: 21-31.
  13. Donoho, D.L., 2006. For most large underdetermined systems of linear equations theminimal 11-norm near-solution is also the sparsest solution. Communic. Pure Appl.Mathemat., 59: 797-829.
  14. Easley, G.R., Labate, D. and Lim, W.Q., 2008. Sparse directional image representationsusing the discrete shearlet transform. Appl. Computat. Harm. Analys., 25: 25-46.
  15. Gaci, S., 2014. The use of wavelet-based denoising techniques to enhance thefirst-arrival picking on seismic traces. IEEE Transact. Geosci. Remote Sens., 58:4558-4563.
  16. Gan, S., Wang, S., Chen, Y., Zhang, Y., and Jin, Z., 2015. Dealiased seismic datainterpolation using seislet transform with low-frequency constraint. IEEE Geosci.Remote Sens. Lett., 12: 2150-2154.
  17. Giiltinay, N., 2003. Seismic trace interpolation in the Fourier transform domain.Geophysics, 68: 355-369.
  18. Hagen, D.C., 1982. The application of principal components analysis to seismic datasets. Geoexplor., 20: 93-111.
  19. Hauser, S. and Steidl, G., 2013. Convex multiclass segmentation with shearletregularization. Internat. J. Comput. Mathemat., 90: 62-81.
  20. Hou, S., Zhang, F., Li, X., Zhao, Q. and Dai, H., 2018. Simultaneous multi-componentseismic denoising and reconstruction via K-SVD. J. Geophys. Engineer., 15: 681.
  21. Hunt, L., Downton, J., Reynolds, S., Hadley, S., Trad, D. and Hadley, M., 2010. Theeffect of interpolation on imaging and AVO: A Viking case study. Geophysics, 75(6):WB265-WB274.
  22. Karbalaali, H., Javaherian, A., Dahlke, S. and Torabi, S., 2017. Channel boundarydetection based on 2D shearlet transformation: An application to the seismic data inthe South Caspian Sea. J. Appl. Geophys., 146: 67-79.
  23. Kong, D. and Peng, Z., 2015. Seismic random noise attenuation using shearlet andtotal generalized variation. J. Geophys. Engineer., 12: 1024.
  24. Li, Q. and Gao, J., 2013. Contourlet based seismic reflection data non-local noisesuppression. J. Appl. Geophys., 95: 16-22.
  25. Liang, J., Ma, J. and Zhang, X., 2014. Seismic data restoration via data-driven tightframe. Geophysics, 79(3): V65-V74.
  26. Liu, Y., Liu, C. and Wang, D., 2008. A 1D time-varying median filter for seismicrandom, spike-like noise elimination. Geophysics, 74(1): V17-V24.
  27. Liu, Y. and Fomel, S., 2013. Seismic data analysis using local time-frequencydecomposition. Geophys. Prosp., 61: 516-525.
  28. Liu, W., Cao, S. and Chen, Y., 2016a. Seismic time-frequency analysis via empiricalwavelet transform. IEEE Geosci. Remote Sens. Lett., 13: 28-32.
  29. Liu, W., Cao, S., Chen, Y. and Zu, S., 2016b. An effective approach to attenuaterandom noise based on compressive sensing and curvelet transform. J. Geophys.Engineer., 13: 135.
  30. Liu, J., Chou, Y. and Zhu, J., 2018. Interpolating seismic data via the POCS methodbased on shearlet transform. J. Geophys. Engineer., 15: 852-876.
  31. Mousavi, S.M. and Langston, C.A., 2016. Hybrid seismic denoising usinghigher-order statistics and improved wavelet block thresholding. Bull. Seismol.Soc. Am., 106: 1380-1393.
  32. Nalla, P.R. and Chalavadi, K.M., 2015. Iris classification based on sparserepresentations using on-line dictionary learning for large-scale de-duplicationapplications. Springer Plus, 4: 238.
  33. Ophir, B., Lustig, M. and Elad, M., 2011. Multi-scale dictionary learning usingwavelets. IEEE J. Select. Topics Sign. Process., 5: 1014-1024.
  34. Ramaswami, S., Kawaguchi, Y., Takashima, R., Endo, T. and Togami, M., 2017.
  35. ADMM-based audio reconstruction for low-cost-sound-monitoring. 25th Europ.Sign.
  36. Process. Conf, (EUSIPCO). doi:10.23919/EUSIPCO.2017.8081410|
  37. Rubinstein, R., Zibulevsky, M. and Elad, M., 2010. Double sparsity: Learning sparsedictionaries for sparse signal approximation. IEEE Transact. Sign. Process., 58:1553-1564.
  38. Sacchi, M.D. and Liu, B., 2005. Minimum weighted norm wavefield reconstructionfor AVA imaging. Geophys. Prosp., 53: 787-801.
  39. Siahsar, M.A.N., Gholtashi, S., Torshizi, E.O., Chen, W. and Chen, Y., 2017.
  40. Simultaneous denoising and interpolation of 3D seismic data via damped data-drivenoptimal singular value shrinkage. IEEE Geosci. Remote Sens. Lett., 14: 1086-1090.
  41. Tang, G., Ma, J.W. and Yang, H.Z., 2012. Seismic data denoising based onlearning-type overcomplete dictionaries. Appl. Geophys., 9: 27-32.
  42. Tong, Q., Sun, Z., Nie, Z., Lin, Y. and Cao, J., 2016. Sparse decomposition based on
  43. ADMM dictionary learning for fault feature extraction of rolling element bearing. J.Vibroengineer., 18: 5204-5216.
  44. Ursin, B. and Zheng, Y., 1985. Identification of seismic reflections using singular valuedecomposition. Geophys. Prosp., 33: 773-799.
  45. Vassiliou, A.A. and Garossino, P., 1998. Time-frequency processing and analysis ofseismic data using very short-time Fourier transforms. U.S. Patent 5,850,622.
  46. Wang, Z., Zhang, B., Gao, J., Wang, Q. and Liu H., 2017. Wavelet transform withgeneralized beta wavelets for seismic time-frequency analysis. Geophysics, 82(4):047-056.
  47. Wright, J., Yang, A.Y., Ganesh, A., Sastry, S.S. and Ma, Y., 2009. Robust facerecognition via sparse representation. IEEE Transact. Pattern Analys. Mach.Intellig., 31: 210-227.灰 小 偷 12/18/19 2:00 PMComment [1]: | am so sorry that | can not findthe IEEE title and the volume. | find thisreference in this website:https://ieeexplore.ieee.org/document/808141And | use this website's citation function. lsthis format right?
  48. Wu, L. and Castagna, J.P., 2017. S-transform and Fourier transform frequency spectra ofbroadband seismic signals. Geophysics, 82(5): 071-081.
  49. Xu, S., Zhang, Y. and Lambaré G., 2010. Antileakage Fourier transform for seismicdata regularization in higher dimensions. Geophysics, 75(6): WB113-WB120.
  50. Xue, Y., Man, M., Zu, S., Chang, F. and Chen, Y., 2017. Amplitude-preserving iterativedeblending of simultaneous source seismic data using high-order Radon transform. J.Appl. Geophys., 139: 79-90.
  51. Yang, H., Long, Y., Lin, J., Zhang, F. and Chen, Z., 2017. A seismic interpolation anddenoising method with curvelet transform matching filter. Acta Geophys., 65:1029-1042.
  52. Yi, S., Labate, D., Easley, G.R. and Krim, H., 2009. A shearlet approach to edgeanalysis and detection. IEEE Transact. Image Process., 18: 929-941.
  53. Yu, S., Ma, J., Zhang, X. and Sacchi, M.D., 2015. Interpolation and denoising ofhigh-dimensional seismic data by learning a tight frame. Geophysics, 80(5):V119-V132.
  54. Yu, S., Ma, J. and Osher, S., 2016. Monte Carlo data-driven tight frame for seismic datarecovery. Geophysics, 81(4): V327-V340.
  55. Zhao, Q. and Du, Q., 2017. Constrained data-driven tight frame for robust seismic datareconstruction. Expanded Abstr., 87th Ann. Internat. SEG Mtg., Houston: 4246-4250.
  56. Zhao, X., Li, Y., Zhuang, G., Zhang, C. and Han, X., 2016. 2-D TFPF based on
  57. Contourlet transform for seismic random noise attenuation. J. Appl. Geophys., 129:158-166.
  58. Zhu, L., Liu, E. and McClellan, J.H., 2015. Seismic data denoising through multiscaleand sparsity-promoting dictionary learning. Geophysics, 80(6): WD45-WD57.
  59. Zhuang, G., Li, Y., Liu, Y., Lin, H., Ma, H., and Wu, N., 2014. Varying-window-length
  60. TFPF in high-resolution Radon domain for seismic random noise attenuation. IEEEGeosci. Remote Sens. Lett., 12: 404-408.
  61. Zou, H., Hastie, T. and Tibshirani, R., 2006. Sparse principal component analysis. J.Computat. Graphic. Statist., 15: 265-286.
Share
Back to top
Journal of Seismic Exploration, Electronic ISSN: 0963-0651 Print ISSN: 0963-0651, Published by AccScience Publishing