中文摘要：

预兆的 deconvolution 算法(PD ) 基于秒顺序统计，假设 primaries 和 multiples 是含蓄地直角的。然而，地震数据通常不在实践满足这个假设。自从地震数据(primaries 和 multiples ) ，有 non-Gaussian 分布，在这篇论文我们在场由最大化恢复 primaries 的 non-Gaussianity 的一个改进预兆的 deconvolution 算法(IPD ) 。合成、真实的地震数据集上的 IPD 方法的应用证明建议方法获得有希望的结果。

英文摘要：

The predictive deconvolution algorithm （PD）, which is based on second-order statistics, assumes that the primaries and the multiples are implicitly orthogonal. However, the seismic data usually do not satisfy this assumption in practice. Since the seismic data （primaries and multiples） have a non-Gaussian distribution, in this paper we present an improved predictive deconvolution algorithm （IPD） by maximizing the non-Gaussianity of the recovered primaries. Applications of the IPD method on synthetic and real seismic datasets show that the proposed method obtains promising results.