A Seislet U-Net for seismic data reconstruction

Abstract

Field seismic data often suffer from trace missing due to acquisition constraints. While deep learning has advanced reconstruction tasks, existing models often lack physical interpretability, risking geologically implausible results. To bridge this gap, we propose Seislet U-Net, which integrates forward and inverse Seislet transforms into a U-Net encoder-decoder. By leveraging local slope information to guide predictions along dominant seismic dips, this design enforces feature extraction and reconstruction in a physically consistent, sparse domain that aligns with seismic wavefront structures, thereby enhancing interpretability. Training uses a composite loss that balances sparsity, spatial smoothness, and structural fidelity. Experiments on synthetic and field datasets demonstrate that Seislet U-Net outperforms U-Net, denoising convolutional neural network, discrete wavelet transform U-Net, and AU-Net, achieving signal-to-noise ratio improvements of 1.47 dB and 4.49 dB, respectively, compared to U-Net. The framework integrates data-driven learning with domain-specific constraints, offering a reliable solution for seismic reconstruction.

Keywords

Seismic data reconstruction / Deep learning / U-Net / Seislet transforms

References

[1] Wang C, Huang X, Li Y, Jensen K. Removing multiple types of noise of distributed acoustic sensing seismic data using attention—guided denoising convolutional neural network. Front Earth Sci. 2022; 10: 986470. doi: 10.3389/feart.2022.986470

[2] Zhao Y, Li Y, Jia X, Wu X. A deghosting method based on deep learning. In: Journal of Physics: Conference Series. 2023 Conference on Geophysics and Deep Learning; May 12—14, 2023; Beijing, China. IOP Publishing; 2023; 2651(1): 012126. doi: 10.1088/1742—6596/2651/1/012126

[3] Zhu Y, Cao J, Yin H, Zhao J, Gao K. Seismic data reconstruction based on attention U—net and transfer learning. J Appl Geophys. 2023; 219: 105241. doi: 10.1016/j.jappgeo.2023.105241

[4] Sacchi MD, Ulrych TJ, Walker CJ. Interpolation and extrapolation using a high—resolution discrete Fourier transform. IEEE Trans Signal Process. 1998; 46(1): 31-38. doi: 10.1109/78.651165

[5] Cao S, Chen X. The second—generation wavelet transform and its application in denoising of seismic data. Appl Geophys. 2005; 2(2): 70-74. doi: 10.1007/s11770—005—0034—4

[6] Thorson JR, Claerbout JF. Velocity—stack and slant—stack stochastic inversion. Geophysics. 1985; 50(12): 2727-2741. doi: 10.1190/1.1441893

[7] Neelamani R, Baumstein AI, Gillard DG, Hadidi MT, Soroka WL. Coherent and random noise attenuation using the curvelet transform. Lead Edge. 2008; 27(2): 240-248. doi: 10.1190/1.2840373

[8] Fomel S, Liu Y. Seislet transform and seislet frame. Geophysics. 2010; 75(3): V25-V38. doi: 10.1190/1.3380591

[9] Kong D, Peng Z. Seismic random noise attenuation using shearlet and total generalized variation. J Geophys Eng. 2015; 12(6): 1024-1035. doi: 10.1088/1742—2132/12/6/1024

[10] Abma R, Kabir N. 3D interpolation of irregular data with a POCS algorithm. Geophysics. 2006; 71(6): E91-E97. doi: 10.1190/1.2356088

[11] Ronneberger O, Fischer P, Brox T. U—Net: convolutional networks for biomedical image segmentation. In: Navab N, Hornegger J, Wells WM, eds. Medical Image Computing and Computer—Assisted Intervention — MICCAI 2015. 18th International Conference; October 5—9, 2015; Munich, Germany. Springer; 2015: 234-241. doi: 10.1007/978—3—319—24574—4_28

[12] Dou Y, Li K, Duan H, Li T, Dong L, Huang Z. MDA GAN: Adversarial—learning—based 3—D seismic data interpolation and reconstruction for complex missing. IEEE Trans Geosci Remote Sens. 2023; 61: 5905014. doi: 10.1109/TGRS.2023.3249476

[13] Liu B, Dong X, Xu L, Qin B. Seismic data reconstruction method using generative adversarial network based on moment reconstruction error constraint. Sci Prog. 2024; 107(2). doi: 10.1177/00368504231208497

[14] Zhang Y, Zhang Y, Dong H, Song L. An integrated method of seismic data reconstruction and denoising based on generative adversarial network. Oil Geophys Prospect. 2024; 59(4): 714-723. [In Chinese]. doi: 10.13810/j.cnki.issn.1000—7210.2024.04.008

[15] Yu W, Dong S, Lu S, Dong X. A novel strategy for simultaneous super—resolution reconstruction and denoising of post—stack seismic profile. Geophys Prospect. 2025; 73(1): 96-112. doi: 10.1111/1365—2478.13646

[16] Dong X, Wei C, Zhong T, Cheng M, Dong S, Li F. Seismic data reconstruction based on a multicascade self—guided network. Geophysics . 2024; 89(3): V179-V195. doi: 10.1190/geo2022—0712.1

[17] Dong X, Wei C, Lin J, Lu S, Cheng M, Zhang R. Deep—learning—based Global Feature Capture for Seismic Data Reconstruction. Geophysics. 2025; 90(5): V499-V511. doi: 10.1190/geo2024—0445.1

[18] Liu N, Liu K, Wu L, Wu H, Zhang Y, Gao J. Polarity—constrained dual—cycle generative adversarial networks for irregularly sampled seismic data reconstruction. IEEE Trans Geosci Remote Sens . 2024; 62: 5918410. doi: 10.1109/TGRS.2024.3411782

[19] Wang Y, Xu X, Wang X. Reconstruction and quality control of non—repeatable land time—lapse seismic data. J Seism Explor. 2025; 34(3): 76-92. doi: 10.36922/JSE025230010

[20] Wu H, Feng J, Liu X, et al. Seismic Impedance Inversion Using Conditional Generative Adversarial Networks Assisted by Spectral Reconstruction—based Data Augmentation. IEEE Trans Geosci Remote Sens. 2025; 63: 5923712. doi: 10.1109/TGRS.2025.3621119

[21] Yu J, Wu B. Attention and hybrid loss guided deep learning for consecutively missing seismic data reconstruction. IEEE Trans Geosci Remote Sens. 2022; 60: 5902108. doi: 10.1109/TGRS.2021.3068279

[22] Huang H, Wang T, Cheng J, Xiong Y, Wang C, Geng J. Self—supervised deep learning to reconstruct seismic data with consecutively missing traces. IEEE Trans Geosci Remote Sens. 2022; 60: 5911514. doi: 10.1109/TGRS.2022.3148994

[23] Liu N, Wu L, Wang J, Wu H, Gao J, Wang D. Seismic data reconstruction via wavelet—based residual deep learning. IEEE Trans Geosci Remote Sens . 2022; 60: 4508213. doi: 10.1109/TGRS.2022.3152984

[24] Chen X, Chen H, Hu Y, Xie Y, Wang S. A sparse time—frequency reconstruction approach from the synchroextracting domain. Signal Process. 2024; 222: 109517. doi: 10.1016/j.sigpro.2024.109517

[25] Saad OM, Ravasi M, Alkhalifah T. Noise attenuation in distributed acoustic sensing data using a guided unsupervised deep learning network. Geophysics. 2024; 89(6): V573-V587. doi: 10.1190/geo2024—0109.1

[26] Li C, Saad OM, Chen Y. Unsupervised deep learning for off—the—grid seismic reconstruction and denoising. Geophysics. 2025; 90(3): V241-V254. doi: 10.1190/geo2024—0189.1

[27] Cao W, Lu W, Li Y. Semi—Supervised W—KAN for Regularly Missing Seismic Data Reconstruction. IEEE Geosci Remote Sens Lett. 2025; 22: 7507405. doi: 10.1109/LGRS.2025.3583567

[28] Peng T, Liu Y, Liu C, Chen C, Chen J. Seislet weighted stacking method with high resolution for low SNR seismic data. Chin J Geophys. 2025; 68(3): 1087-1101. [In Chinese]. doi: 10.6038/cjg2024R0866

[29] Zhao H, Gallo O, Frosio I, Kautz J. Loss functions for image restoration with neural networks. IEEE Trans Comput Imaging. 2017; 3(1): 47-57. doi: 10.1109/TCI.2016.2644865

[30] Chen Y, Ma J, Fomel S. Double—sparsity dictionary for seismic noise attenuation. Geophysics . 2016; 81(2): V103-V116. doi: 10.1190/geo2014—0525.1

[31] SEG Wiki. SEG/EAGE 3D modeling Salt Model Phase—C 1996 . SEG Wiki. Updated May 2, 2016. Available from: https://wiki.seg.org/wiki/SEG/EAGE_3D_modeling_Salt_Model_Phase—C_1996 [Last accessed on October 5, 2023].

[32] Zhang K, Zuo W, Chen Y, Meng D, Zhang L. Beyond a Gaussian denoiser: Residual learning of deep CNN for image denoising. IEEE Trans Image Process. 2017; 26(7): 3142-3155. doi: 10.1109/TIP.2017.2662206

[33] Lou Y, Wu L, Liu L, et al. Irregularly sampled seismic data interpolation via wavelet—based convolutional block attention deep learning. Artif Intell Geosci. 2022; 3: 192-202. doi: 10.1016/j.aiig.2022.12.001

[34] SEG Wiki. PGS Simultaneous Source Marine Line . SEG Wiki. Updated May 17, 2018. Available from: https://wiki.seg.org/wiki/PGS_Simultaneous_Source_Marine_Line [Last accessed on October 5, 2023].