稀疏反演多次波去除策略与效果分析

多次波去除是海上勘探中的一项关键技术，在一定程度上决定着最终成像结果的质量.表面相关多次波去除方法（Surface-Related Multiple Elimination，SRME）是一种基于反馈迭代模型的数据驱动类方法，该方法利用地震数据本身预测多次波，进一步对其进行匹配相减获得一次波结果，这往往导致与多次波重叠的一次波被错误地去除.稀疏反演一次波估计方法（Estimation of Primaries by Sparsity Inversion，EPSI）在原理上同样基于反馈迭代模型，通过求解一次波和表面相关多次波总残差最小化问题直接获得一次波.该方法无多次波的匹配相减过程，可减少与多次波发生重叠的一次波产生损失.另外，EPSI可以很好地解决SRME无法解决的近道缺失问题.理论数据和实际资料处理对比表明SRME对数据完整性以及一次波和多次波的分布情况有较强的依赖性，而EPSI几乎不受这两个因素的影响.EPSI相比SRME方法计算成本高，尚未在工业界广泛应用，但随着高性能计算的普及，该方法的实用性有望得到突破.

Strategy of multiple elimination by sparsity inversion and its effectiveness analysis

Multiple elimination is one of the most important technique in marine exploration. Surface-Related Multiple Elimination predicts multiples and subtracts them from the full wavefield to obtain the primaries, which may result in the distortion of the primaries. The Estimation of Primaries by Sparsity Inversion method minimizes the estimation residual of the full waveform adding the sparsity constraint of primary impulse responses. The primaries are estimated directly through corresponding iterative inversion process, which relaxes the distortion of primaries when primaries and multiples overlap. Furthermore, EPSI can overcome the near-offset missing problem through estimating primaries from the multiples. The synthetic and field data tests demonstrated that EPSI is less effected by the data-integrity and the distribution of the primaries and the multiples than SRME. EPSI has high calculation costs and thus is not widely applied in industry, yet it will become more practical with the development of pervasive high-performance computing technology.