基于曲波域稀疏约束的OVT域地震数据去噪方法研究  

A seismic denoising method based on curvelet transform with sparse constraint in VT domain  

（）1.中国石油大学(华东)地球科学与技术学院,山东青岛266580;2.中石化石油工程地球物理有限公司胜利分公司,山东东营257088;3.中国石油化工股份有限公司胜利油田分公司物探研究院,山东东营257000;4.中国石油集团东方地球物理勘探有限责任公司,河北涿州072751)  

（1.School of Geosciences,China University of Petroleum,Qingdao 266580,China;2.Shengli Branch Company,SGC,Dongying 257088,China;3.Geophysical Research
Institute,Shengli Oilfield,Sinopec,Dongying 257000,China;4.BGP INC.,China National Petroleum Corporation,Zhuozhou 072751,China)  

炮检距向量片(offset vector tile,OVT)道集中噪声与有效信号的差异小,深层的弱有效信号同相轴连续性差,传统去噪方法在抑制噪声的同时会对弱有效信号造成较大损伤。为解决这一问题,在OVT道集中引入了基于压缩感知(compressed sensing,CS)理论的曲波域稀疏约束地震数据去噪方法,该方法基于曲波变换的多方向性和各向异性对地震数据进行稀疏描述,利用与噪声相关的信息约束正交匹配追踪(orthogonal matching pursuit,OMP)重构算法的迭代过程,实现对弱有效信号的提取。模型测试和实际资料处理结果表明:小波阈值去噪方法在抑制噪声的同时会损伤与噪声差异小的弱有效信号,对同相轴的连续性改善不明显,造成深层弱有效信号的同相轴连续性差;CS小波去噪方法可一定程度保护弱有效信息,但由于无法精确表达直线或曲线等边缘特征,分离与噪声差异小的深层弱信号及噪声时效果不理想;基于CS理论的曲波域稀疏约束地震数据去噪方法克服了OMP重构算法对信号稀疏度的依赖,有效提取了OVT域地震数据的中、深反射层的弱有效信号,在压制强随机噪声的同时减少了弱有效信号的损失,提高了地震剖面的信噪比和同相轴的连续性。

 

In offset vector tile (OVT) gather,a minor difference exists between noise and the effective signals,where weak effective signals from deep layers show poor event continuity.Traditional denoising methods cause considerable damage to weak effective signals when denoising.In this study,a sparse constraint denoising algorithm in the curvelet domain is proposed based on compressed sensing (CS) and is applied to the denoising of OVT gathers with a low signal-to-noise ratio (SNR).First,the seismic data are represented in the curvelet domain.Then,noise-related information is used to constrain the iterations of the orthogonal matching pursuit (OMP) reconstruction algorithm.Finally,the weak effective signals are reconstructed by the inverse curvelet transform.Tests on both model and field data demonstrate that the wavelet threshold denoising method can damage weak effective signals while suppressing noise.And,the method does not improve the continuity of events,particularly weak signals from deep layers.The CS wavelet denoising method can protect weak effective information to some extent,but it cannot effectively extract weak signals,which are slightly different to noise.This is because the wavelet transform cannot precisely express edge features such as straight lines or curves.The proposed method overcomes the dependence of the OMP algorithm on signal sparsity and effectively extracts weak signals from deep layers of the actual data in the OVT domain while simultaneously improving both the SNR and the continuity of seismic events.

国家重点研发计划(2016YFC060110501)、国家重点基础研究发展计划(973计划)项目(2014CB239006)和国家科技重大专项(2017ZX05005004-03,2016ZX05006-002-003,2016ZX05026-002-002)共同资助。

 

10.3969/j.issn.1000-1441.2019.02.006