基于双重稀疏表示的地震资料随机噪声衰减方法  

Seismic random seismic noise attenuation method on basis of the double sparse representation

作　　者：罗勇[1] 毛海波[1] 杨晓海[1] 李文捷[1] 陈文超[2] LUO Yong;MAO Hai-Bo;YANG Xiao-Hai;LI Wen-Jie;CHEN Wen-Chao(Institute of Geophysics,Research Institute of Petroleum Exploration ＆ Development,Urumqi 830013,China;School of Electronic ＆ Information Engi？neering,Xi＇an Jiaotong University,Xi＇an 710049,China)

机构地区：[1]新疆油田分公司勘探开发研究院地球物理研究所,新疆乌鲁木齐830013 [2]西安交通大学电子与信息工程学院,陕西西安710049

出　　处：《物探与化探》2018年第3期608-615,共8页Geophysical and Geochemical Exploration

基　　金：国家自然科学基金项目(41774135,41504092,41274125);中国博士后科学基金项目(2016T90925,2015M572567);中央高校基本科研业务费专项资金资助

摘　　要：针对固定字典难以完全匹配实际资料复杂的形态特征,以及学习字典不具备快速算法、计算复杂等问题,文中选择双重稀疏字典来兼备结构性和自适应性,不仅降低了训练样本的数量,而且更适于高维信号的分析。该方法以过完备离散余弦变换(overcomplete discrete cosine transform,ODCT)作为训练基字典,将待处理资料的特征数据作为样本,利用稀疏K-SVD算法,建立了基于双重稀疏字典的地震随机噪声衰减模型。典型的合成及实际高维地震资料处理结果表明,本文方法不仅可以有效地对地震资料随机噪声进行衰减,而且能更好地保持断层等边缘结构。The double sparse dictionary is adopted for the seismic random noise attenuation. The seismic data are not represented well by the fixed dictionaries,which do not contain the effective information about the seismic data; the learning dictionaries are fully adaptable but are costly to deploy in the big data processing.The double sparse dictionary reduces the number of training sample and is more suitable for the construction of high-dimension dictionary and the analysis of the high-dimension signal. With the over completed discrete cosine transform as the base dictionary,the sparse dictionary is trained by the sparse K-SVD driven by the noisy seismic data samples.Thus the seismic random noise attenuation model based on the double sparse dictionary is established.A comparison of the results of the synthesized and real data in high dimensional case shows that the seismic random noise can be suppressed effectively by the method based on double sparse dictionary and the fault structure can be preserved in 3 D case.

关 键 词：噪声衰减 稀疏表示 学习字典 形态成分分析 稀疏K-SVD 

分 类 号：P631.4[天文地球—地质矿产勘探]