基于独立分量分析基的地震随机噪声压制

传统的独立分量分析(Independent Component Analysis,ICA)去噪方法假设地震记录的相邻道含有相同的随机噪声,仅适用于同相轴较平的地震记录,去噪效果并不显著。为了改善ICA方法对高斯随机噪声的压制效果,首先通过构造度量数据非高斯性的目标函数求取地震数据的ICA基,将数据转换至ICA域;然后通过贝叶斯方法构造出满足非高斯分布的阈值函数,进行阈值法去噪处理。为了满足独立分量分析的假设条件,将地震数据进行分块处理,并假设每个数据块与整体的数据含有相似的数据结构。理论模型及实际资料试算结果表明,该方法可以有效地压制剖面中的高斯随机噪声,对含复杂界面的数据也十分有效,具有较好的应用价值。

Traditional seismic denoising methods based on independent component analysis (ICA) assume that adjacent traces in seismic record contain the same random noise.The denoising effect based on this assumption is not significant,and it is only suitable for seismic records with flat event.Therefore,we convert the seismic data to ICA domain by constructing an ICA basis according to the objective function measuring the non-Gaussianity and set a threshold to reduce the random noise.This threshold function is obtained by Bayesian method.In order to satisfy the assumption of independent component analysis,we carried out the partition processing on the seismic data and supposed that each data block has the similar structures with the whole data.The processing results of theoretical and real data show that this method can suppress the random noise effectively and it is suitable for seismic data with complex interface.So the denoising method based on independent component analysis has the obvious advantage and good application value for Gaussian random noise suppression.

国家自然科学基金(41374123)和国家科技重大专项(2011ZX05006-002)项目共同资助。