U-STDRNet: A unified model integrating swin transformer and residual dense network for seismic image super-resolution and denoising

Abstract

Enhancing seismic image resolution while effectively suppressing noise remains a critical challenge in accurately characterizing subsurface geological structures for oil and gas exploration. Traditional methods often fail to balance the recovery of fine details with robustness to noise, particularly in complex geological settings or under high-noise conditions. This study proposes a deep learning-based joint model, U-Net Shifted Window (Swin) Transformer-based dense residual network (U-STDRNet). The model integrates the global modeling capability of the Swin Transformer, the hierarchical feature reuse mechanism of the residual dense network, and an attention-guided strategy to jointly perform seismic image super-resolution and denoising. Built upon the U-Net encoder-decoder architecture, the model embeds Swin Transformer-based convolutional residual blocks. These blocks employ both a feature fusion block with the Swin Transformer and a feature fusion block with a convolutional neural network to effectively capture stratigraphic continuity and enhance detailed features such as fault edges. Residual dense blocks further improve weak signal recovery (e.g., thin-layer interfaces) through dense residual connections. Furthermore, the convolutional block attention module is integrated into skip connections, employing a dual-channel spatial weighting mechanism to suppress noise and emphasize key geological regions. Experimental results and field-data experiments demonstrate that U-STDRNet achieves a higher peak signal-to-noise ratio than the traditional U-Net. In addition, the model successfully restores fault and fold continuity details while exhibiting superior noise suppression compared to existing methods.

Keywords

U-Net Swin Transformer-based dense residual network / Seismic image super-resolution / Seismic image denoising / Deep learning models / Swin transformer / Residual dense networks / Convolutional block attention module

Acknowledgments

None.

Funding

This work was supported in part by the National Key Research and Development Program of China under Grant 2024YFB4007100, in part by the National Major Science and Technology Projects of China under Grant 2024ZD1004300, in part by National Natural Science Foundation of China under Grant 42304133 and 42574175, in part by Key Project of the Education Department of Hubei Province (Grant No. D20241304), and in part by Key project from the Hubei Research Center for Basic Disciplines of Earth Sciences under Grant HRCES-202401.

Conflict of interest

The authors declare no conflicts of interest.

Author contributions

Conceptualization: Min Bai, Guangtan Huang

Formal analysis: Juan Wu

Investigation: Mingliao Wu, Juan Wu, Haiyu Li

Methodology: Juan Wu, Zhixian Gui

Writing–original draft: Mingliao Wu

Writing–review & editing: Mingliao Wu, Guangtan Huang

Availability of data

The data generated or analyzed during this study are available from the corresponding author upon reasonable request.

References

[1] Bianco MJ, Gerstoft P, Olsen KB, Lin FC. High-resolution seismic tomography of long beach, CA using machine learning. Sci Rep. 2019; 9:14987. doi: 10.1038/s41598-019-50381-z

[2] Zhang Y, Wang Y, Yin J. Single point high density seismic data processing analysis and initial evaluation. Oil Geophys Prospect. 2010; 45(2):201-207. doi: 10.13810/j.cnki.issn.1000-7210.2010.02.008

[3] Rebert T, Sablon R, Vidal N, Charrier P, Soubaras R. Improving pre-salt imaging with variable-depth streamer data. In: SEG Technical Program Expanded Abstracts 2012. United States: SEG; 2012. p. 1-5. doi: 10.1190/segam2012-1067.1

[4] Wu X, Gan L, Yuan S, Rui D. A preliminary study on wellbore flow interpretation of fiber optic vibration signals based on K-means clustering algorithm. SN Appl Sci. 2022; 4(8):233. doi: 10.1007/s42452-022-05117-6

[5] Wang Y, Wang J, Wang X, Sun W, Zhang J. Broadband Processing Key Technology Research and Application on Slant Streamer. In: International Geophysical Conference, Beijing, China, 24-27 April 2018. United States: SEG; 2018. p. 135-138. doi: 10.1190/IGC2018-034

[6] Mu XR, Mao Q, Huang JP. Stable attenuation-compensated reverse time migration and its application to land seismic data. Petrol Sci. 2023; 20(5):2784-2795. doi: 10.1016/j.petsci.2023.03.014

[7] Wang Z, Yue WZ, Zhu YM, Ji NX, Fan SS. Near-wellbore 3D velocity imaging inversion method based on array acoustic logging data. Petrol Sci. 2025; 22:002341. doi: 10.1016/j.petsci.2024.06.007

[8] Wang C, Chen L, Li L, et al. Combining unscented Kalman filter and wavelet neural network for anti-slug. Petrol Sci. 2023; 20(6):3752-3765. doi: 10.1016/j.petsci.2023.05.008

[9] Li L, Wang ZZ, Yin SD, et al. Selection and application of wavelet transform in high-frequency sequence stratigraphy analysis of coarse-grained sediment in rift basin. Petrol Sci. 2024; 21(5):3016-3028. doi: 10.1016/j.petsci.2024.06.020

[10] Liao QZ, Xue L, Lei G, Liu X, Sun SY, Patil S. Statistical prediction of waterflooding performance by K-means clustering and empirical modeling. Petrol Sci. 2022; 19(3):1139-1152. doi: 10.1016/j.petsci.2021.12.032

[11] Bai M, Huang G, Wang H, Chen Y. Seismic signal enhancement based on the low-rank methods. Geophys Prospect. 2020; 68(9):2783-2807. doi: 10.1111/1365-2478.13028

[12] Huang G, Bai M, Zhao Q, Wang H, Chen W, Chen Y. Erratic noise suppression using iterative structure-oriented space-varying median filtering with sparsity constraint. Geophys Prospect. 2021; 69(1):101-121. doi: 10.1111/1365-2478.13032

[13] Wu J, Bai M, Zhang D, Wang H, Huang G, Chen Y. Fast and robust low-rank approximation for five-dimensional seismic data reconstruction. IEEE Access. 2020; 8:175501-17512. doi: 10.1109/ACCESS.2020.3026020

[14] Liu W, Cheng Q, Liu L, Wang Y, Zhang J. Accelerating high-resolution seismic imaging by using deep learning. Appl Sci Basel. 2020; 10(7):2502. doi: 10.3390/app10072502

[15] Yu S, Ma J, Wang W. Deep learning for denoising. Geophysics. 2019; 84(6):V333-V350. doi: 10.1190/geo2018-0668.1

[16] Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation; Proceedings of the Medical Image Computing and Computer-Assisted Intervention - MICCAI 2015. Cham, F: Springer International Publishing; 2015. doi: 10.1007/978-3-319-24574-4_28

[17] Wang Y, Lu W, Liu J, Zhang M, Miao Y. Random seismic noise attenuation based on data augmentation and CNN. Chin J Geophys Chin. 2019; 62(1):421-433. doi: 10.6038/cjg2019M0385

[18] Dong X, Zhong T, Wang H, Wu N, Li Y, Yang B. The denoising of desert seismic data acquired from tarim basin based on convolutional adversarial denoising network. Chin J Geophys. 2022; 65(7):2661-2672. doi: 10.6038/cjg2022P0279

[19] Bai T, Zhao H, Wang Z. A U-Net Based Deep Learning Approach for Seismic Random Noise Suppression; Proceedings of the IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Symposium; 2022. doi: 10.1109/IGARSS46834.2022.9880193

[20] Halpert AD. Deep learning-enabled seismic image enhancement. In: SEG Technical Program Expanded Abstracts 2018. United States: SEG; 2018. p. 2081-2085. doi: 10.1190/segam2018-2996943.1

[21] Jiang J, Ren H, Zhang M. A convolutional autoencoder method for simultaneous seismic data reconstruction and denoising. IEEE Geosci Remote Sens Lett. 2022; 19:1-5. doi: 10.1109/LGRS.2021.3073560

[22] Ma H, Yao H, Li Y, Wang H. Deep residual encoder-decoder networks for desert seismic noise suppression. IEEE Geosci Remote Sens Lett. 2020; 17(3):529-533. doi: 10.1109/LGRS.2019.2925062

[23] Wang F, Chen S. Residual learning of deep convolutional neural network for seismic random noise attenuation. IEEE Geosci Remote Sens Lett. 2019; 16(8):1314-1318. doi: 10.1109/LGRS.2019.2895702

[24] Zhao Z, Rao Y. Seismic random noise suppression based on SP-DnCNN neural network. Chin J Geophys Chin. 2024; 67(10):3841-3850. doi: 10.6038/cjg2023R0483

[25] Zhu W, Mousavi SM, Beroza GC. Seismic signal denoising and decomposition using deep neural networks. IEEE Trans Geosci Remote Sens. 2019; 57(11):9476-9488. doi: 10.1109/TGRS.2019.2926772

[26] Zhang S, Zhang L, Qin X. Seismic signal denoising using variational mode decomposition and a denoising convolutional neural network. J Seim Explor. 2025; 34(2):025260030. doi: 10.36922/JSE025260030

[27] Liu N, Wang J, Gao J, Chang S, Lou Y. Similarity-informed self-learning and its application on seismic image denoising. IEEE Trans Geosci Remote Sens. 2022; 60:1-13. doi: 10.1109/TGRS.2022.3210217

[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 Chin. 2025; 68(3):1087-1101. doi: 10.6038/cjg2024R0866

[29] Chen G, Liu Y, Zhang M, Zhang H. Dropout-based robust self-supervised deep learning for seismic data denoising. IEEE Geosci Remote Sens Lett. 2022; 19:1-5. doi: 10.1109/LGRS.2022.3167999

[30] Liu B, Yue J, Zuo Z, et al. Unsupervised deep learning for random noise attenuation of seismic data. IEEE Geosci Remote Sens Lett. 2022; 19:1-5. doi: 10.1109/LGRS.2021.3057631

[31] Shao D, Zhao Y, Li Y, Li T. Noisy2Noisy: Denoise pre-stack seismic data without paired training data with labels. IEEE Geosci Remote Sens Lett. 2022; 19:1-5. doi: 10.1109/LGRS.2022.3145835

[32] Shao J, Wang Y, Liang X, Xue Q, Liang E, Shi S. Siamese network based noise elimination of artificial seismic data recorded by distributed fiber-optic acoustic sensing. Chin J Geophys Chin. 2022; 65(9):3599-3609. doi: 10.6038/cjg2022P0919

[33] Wang S, Hu W, Yuan P, et al. A self-supervised deep learning method for seismic data deblending using a blind-trace network. IEEE Trans Neural Netw Learn Syst. 2023; 34(7):3405-3414. doi: 10.1109/TNNLS.2022.3188915

[34] Xu Z, Luo Y, Wu B, Meng D, Chen Y. Deep nonlocal regularizer: A self-supervised learning method for 3-D seismic denoising. IEEE Trans Geosci Remote Sens. 2023; 61:1-17. doi: 10.1109/TGRS.2023.3329303

[35] Geetha K, Hota MK, Karras DA. A novel approach for seismic signal denoising using optimized discrete wavelet transform via honey badger optimization algorithm. J Appl Geophys. 2023; 219:105236. doi: 10.1016/j.jappgeo.2023.105236

[36] Yang B, Bai M, Wu J, et al. Seismic data denoising using convolutional sparse coding with an efficient alternating direction multipliers minimization algorithm. J Appl Geophys. 2025; 233:105610. doi: 10.1016/j.jappgeo.2024.105610

[37] Zhang P, Wang Q, Liu Y, Chen C. A combined denoising method for Q-factor compensation of poststack seismic data. J Appl Geophys. 2024; 229:105500. doi: 10.1016/j.jappgeo.2024.105500

[38] Woo S, Park J, Lee JY, Kweon IS. CBAM: Convolutional Block Attention Module. Berlin: Springer; 2018. doi: 10.1007/978-3-030-01234-2_1

[39] Ding M, Zhou Y, Chi Y. Seismic signal denoising using Swin-Conv-UNet. J Appl Geophys. 2024; 223:105355. doi: 10.1016/j.jappgeo.2024.105355

[40] Gao L, Shen H, Min F. Swin transformer for simultaneous denoising and interpolation of seismic data. Comput Geosci. 2024; 183:105510. doi: 10.1016/j.cageo.2023.105510

[41] Ioffe S, Szegedy C. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift . In: International Conference on Machine Learning Research; 2015. doi: 10.48550/arXiv.1502.03167

[42] Shi W, Caballero J, Huszar F, et al. Real-Time Single Image and Video Super-Resolution using an Efficient Sub-Pixel Convolutional Neural Network; Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR); 2016. doi: 10.1109/CVPR.2016.207

[43] Li J, Wu X, Hu Z. Deep learning for simultaneous seismic image super-resolution and denoising. Ieee Trans Geosci Remote Sens. 2021; 3057857. doi: 10.1109/TGRS.2021.3057857

[44] Wu X, Hale D. 3D seismic image processing for faults. Geophysics. 2016; 81(2):IM1-IM11. doi: 10.1190/geo2015-0380.1

[45] Wu X, Liang L, Shi Y, Fomel S. FaultSeg3D: Using synthetic data sets to train an end-to-end convolutional neural network for 3D seismic fault segmentation. Geophysics. 2019; 84(3):IM35-IM45. doi: 10.1190/geo2018-0646.1

[46] Kingma D, Ba J. Adam: A Method for Stochastic Optimization. Berlin: Springer; 2014. doi: 10.48550/arXiv.1412.6980

[47] Huang G, Zhang D, Chen W, Chen Y. Accelerated signal-and-noise orthogonalization. IEEE Trans Geosci Remote Sens. 2022; 60:1-9. doi: 10.1109/TGRS.2021.3054839

[48] Huang G, Bai M, Wang H, Liu X, Chen Y. Frequency-space-dependent smoothing regularized nonstationary predictive filtering. IEEE Trans Geosci Remote Sens. 2022; 60:1-9. doi: 10.1109/TGRS.2021.3064945