空间自适应EPLL低频地震随机噪声降噪方法

doi:10.19665/j.issn1001-2400.2021.06.025

摘要/Abstract

摘要：

期望块对数似然降噪方法从外部数据学习高斯混合模型作为地震信号先验,利用与图像块最匹配的高斯分量对数据块去噪,并与含噪图像加权实现图像重构,能够有效抑制地震图像中的随机噪声,保持图像细节。期望块对数似然算法采用与噪声方差有关的正则化参数,在处理沙漠地震图像的非平稳信号和低频色噪声时很难兼顾弱信号保真与噪声压制。针对期望块对数似然正则化参数不适应非平稳地震信号问题,提出空间自适应期望块对数似然低频随机噪声降噪方法。该方法采用方差归一化对图像平稳化,根据图像块信噪比控制正则化参数,使其随非平稳地震勘探信号强度的时空变化而自适应调整,从而平衡局部细节的保持和非平稳信号全局特征的恢复。此外,在信号重构过程中,通过块信噪比对图像块加权平均,减少信号丢失。笔者采用空间自适应期望块对数似然算法对合成以及实际地震勘探图像去噪。结果表明,该方法可以有效地恢复沙漠地震勘探图像中非平稳信号,抑制低频弱相似随机噪声。

关键词: 沙漠地震图像, 随机噪声压制, 非平稳信号, 块信噪比, 期望块对数似然降噪方法

Abstract:

The expected patch log likelihood (EPLL) frame utilizes a Gaussian mixture model (GMM) learned from external data as signal priors.The EPLL denoises the image patches via their most likely Gaussian component in the GMM and weighted-average the denoised patches and the noisy image to reconstruct the denoised image,leading to asuccessful denoising performance for random noise in the seismic image.Since the regularization parameter is only associated with the noise variance,it is difficult to achieve the balance between weak signal preservation and noise suppression for the desert seismic images containing non-strationary seismic signals and low-frequency colored noise.However,the EPLL is unadaptable to the non-stationary seismic signals in desert seismic images.A spatially adaptive EPLL (SA-EPLL) algorithm is proposed in this paper under the framework of the EPLL.In this method,we stabilize the seismic image with the variance normalization method and construct the patch signal-to-noise ratio (P-SNR) related regularization parameter,so that it can be adaptively adjusted with the spatiotemporally various intensity of the non-stationary seismic signals,allowing the balance between the preservation of local details and the restoration of global features of the non-stationary signals.In addition,in the signal reconstruction process,the P-SNR is used as the weight to weighted-average the denoised image patches,leading to a better denoising performance in less signal loss.The SA-EPLL algorithm is applied to synthetic and field seismic images,with the results showing that the proposed method can effectively restore non-stationary signals and suppress low-frequency random noise with weak similarity in desert seismic images.

Key words: desert seismic image, random noise suppression, non-stationary signal, patch signal-to-noise ratio, EPLL

中图分类号:

TN911.7

林红波,马阳. 空间自适应EPLL低频地震随机噪声降噪方法[J]. 西安电子科技大学学报, 2021, 48(6): 204-211.

LIN Hongbo,MA Yang. Spatially adaptive EPLL denoising for low-frequency seismic random noise[J]. Journal of Xidian University, 2021, 48(6): 204-211.

图/表 6

图1

图2

图3

表1

图4

图5

参考文献 16

[1]	LI G H, LI Y, YANG B G. Seismic Exploration Random Noise on Land:Modeling and Application to Noise Suppression[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(8):4668-4681. doi: 10.1109/TGRS.2017.2697444
[2]	ZHU W, LI X, LIU C, et al. An STFT-LSTM System for P-Wave Identification[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 17(3):519-523. doi: 10.1109/LGRS.8859
[3]	HU Z, GAO J H, LIU N. Separation of Blended Seismic Data Using the Synchrosqueezed Curvelet Transform[J]. IEEE Geoscience and Remote Sensing Letters, 2020, 17(4):711-715. doi: 10.1109/LGRS.8859
[4]	LIU Z G, XU S H, XIAO D S, et al. Extreme Learning Ridgelet Process Neural Network and Application[J]. Journal of University of Electronic Science and Technology of China, 2019, 48(1):110-116
[5]	LIN H, LI Y, MA H, et al. Simultaneous Seismic Random Noise Attenuation and Signal Preservation by Optimal Spatiotemporal TFPF[J]. Journal of Applied Geophysics, 2016, 128:123-130. doi: 10.1016/j.jappgeo.2016.03.013
[6]	张洁, 李月. 平行径向道时频峰值滤波消减地震资料的随机噪声[J]. 吉林大学学报:工学版, 2014, 44(3):882-887.
	ZHANG JIE, LI Yue. Reduction of Random Noise in Seismic Data by Parallel Radial-Trace Time-Frequency Peak Filtering[J]. Journal of Jilin University:Engineering and Technology Edition, 2014, 44(3):882-887.
[7]	AHARON M, ELAD M, BRUCKSTEIN A. KSVD:An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation[J]. IEEE Transactions on signal processing, 2006, 54(11):4311-4322. doi: 10.1109/TSP.2006.881199
[8]	YANG Y, GAO J, WANG Z, et al. 2-D Seismic Random Noise Attenuation via Self-Paced Nonnegative Dictionary Learning[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12(12):5391-5401. doi: 10.1109/JSTARS.4609443
[9]	王勇, 乔倩倩, 杨笑宇等. 结合自适应字典学习的稀疏贝叶斯重构[J]. 西安电子科技大学学报, 2016, 43(4):1-4.
	WANG Yong, QIAO Qianqian, YANG Xiaoyu, et al. Sparse Bayesian Reconstruction Combined with Self-Adaptive Dictionary Learning[J]. Journal of Xidian University, 2016, 43(4):1-4.
[10]	刘帅, 马文萍, 杨淑媛等. 结合分层字典学习和空谱信息的多光谱图像去噪[J]. 西安电子科技大学学报, 2017, 44(4):174-180.
	LIU Shuai, MA Wenping, YANG Shuyuan, et al. Multispectral Imagery Denoising Using Hierarchical Dictionary Learning with Spatial-Spectral Information[J]. Journal of Xidian University, 2017, 44(4):174-180.
[11]	ZORAN D, WEISS Y. From Learning Models of Natural Image Patches to Whole Image Restoration[C]// International Conference on Computer Vision,Piscataway:IEEE, 2011:479-486.
[12]	PAPYAN V, ELAD M. Multi-Scale Patch-Based Image Restoration[J]. IEEE Transactions on image processing, 2016, 25(1):249-261. doi: 10.1109/TIP.2015.2499698
[13]	SULAM J, ELAD M. Expected Patch Log Likelihood with A Sparse Prior[C]// International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition.Heidelberg:Springer, 2015:99-111.
[14]	XI H R, LIN H B, ZHANG D D, et al. Patch Classification Based EPLL with Mixed GMM for Nonstationary Seismic Random Noise Attenuation[J]. Journal of applied Geophysics, 2020, 174:103955. doi: 10.1016/j.jappgeo.2020.103955
[15]	ZHAO Y X, LI Y, DONG X T, et al. Low-Frequency Noise Suppression Method Based on Improved DnCNN in Desert Seismic Data[J]. IEEE Geoscience & Remote Sensing Letters, 2019, 16(5):811-815.
[16]	PANI B, KLEMENC J, NAGODE M. Optimizing the Estimation of A Histogram-Bin Width—Application to the Multivariate Mixture-Model Estimation[J]. Mathematics, 2020, 8(7):1090-1119. doi: 10.3390/math8071090

含噪数据信噪比/dB	BM3D	EPLL	MSEPLL	SA-EPLL
-0.36	8.76/0.012	6.97/0.034	9.68/0.020	13.82/0.007
-3.88	5.94/0.020	5.55/0.048	6.37/0.034	10.64/0.015
-5.82	5.47/0.031	4.73/0.058	4.94/0.048	8.95/0.022
-7.40	3.54/0.045	4.02/0.068	3.70/0.063	7.53/0.030
-10.26	0.88/0.083	2.75/0.091	1.31/0.111	5.01/0.054