基于Contourlet变换的K-L变换地震随机噪声自适应衰减方法

摘要

After the Contourlet transform,Contourlet coefficients of reflection are large and correlative due to the rich texture and edge features of effective signals in seismic data,while the Contourlet coefficients of random noise are small and unrelated.Consid ering the advantage of the classification feature extraction for the K-L transform,we apply the K-L transform to Contourlet-domain coefficients.Maximum likelihood estimation and multiscale noise estimation are adopted to estimate the variance of Contourlet coefficients of effective signals and random noise in seismic records.In order to adaptively determine the number of eigenvectors used for the K-L inverse transform,the variance is applied to define the energy percentage threshold function in the K-L domain.Contourlet coefficients transform are modified.And then Contourlet inverse transform is conducted to suppress random noise.However,numerical simulation and field data processing results show that this method can effectively suppress random noise with high fidelity.%地震资料中有效反射信号具有丰富的纹理及边缘特性,在Contourlet变换域系数较大并具有相关性,而随机噪声均匀分布于Contourlet变换域且系数较小.考虑K-L变换具有分类特征提取的优势,在Contourlet变换基础上应用K-L变换,采用最大似然估计法和多尺度噪声估计法估算地震记录中有效信号及随机噪声的Contourlet系数方差,并将其应用到K-L变换域能量百分比阈值函数的定义中,自适应地确定用于K-L反变换的特征向量,修改Contourlet变换的系数,再进行Contourlet反变换压制随机噪声.数值模拟及实际地震资料去噪效果表明,基于Contourlet变换的K-L变换去噪方法不仅可以有效地压制地震资料中的随机噪声,提高地震资料信噪比,而且具有较好的保真性.