基于构造导向滤波的多道反褶积方法及应用

传统线性反褶积方法无法大幅度提高地震资料的分辨率，因此逐渐被非线性的稀疏反褶积方法替代。然而大多数非线性反褶积方法是基于单道反演的，没有考虑道与道之间的空间结构关系，得到高分辨率的结果容易产生横向不一致或者不连续问题，而且当地震资料品质较差时，单道反演方法的结果很不稳定。为此，提出一种基于数据驱动和反演策略的多道反褶积方法，即从地震数据本身出发，利用局部倾角获取构造信息，构建构造导向滤波算子，再将其作为正则化算子引入稀疏多道反褶积中，通过分裂Bregman优化算法求解得到稀疏反射系数系列，在提高分辨率的同时，改善了传统非线性反褶积结果的空间连续性。模型试验和实际地震资料处理结果验证了该方法的有效性和实用性。

Multichannel deconvolution method based on structure-oriented filtering and its application

Since the traditional linear deconvolution cannot greatly improve the resolution of seismic data,it is gradually replaced by the nonlinear sparse deconvolution. However,most nonlinear deconvolution methods are based on single-trace inversion and do not consider the spatial structure relationship between traces. Thus,the obtained high resolution results are prone to lateral inconsistency or discontinuity. Moreover,when the seismic data quality is poor,the results of single-trace inversion are quite unstable. Therefore,this paper proposes a data-driven multichannel deconvolution method based on inversion strategies. First, it starts from the seismic data and uses the local dip angle to obtain structural information. Then,a structure-oriented filtering operator is constructed,which is introduced into the sparse multichannel deconvolution as a regularization operator. Finally,the sparse reflection coefficient series is obtained by splitting the Bregman optimization algorithm. This method improves not only the resolution of seismic data but also the spatial continuity of traditional nonlinear deconvolution results. Model test and actual seismic data processing results show that the method is effective and practical.