论文详情
随机噪声的存在会降低地震资料的信噪比,影响对有效信号尤其是不连续性信号的分析。尺度不变性噪声估计方法基于峰度值分布不随尺度变化,能够在复杂低噪声数据上较好地估计噪声水平;加权核范数最小化能够根据矩阵奇异值刻画数据差异,通过给定不同的权值,突显数据中重要的信息。为此研究了基于噪声水平估计的加权核范数最小化噪声压制方法,利用尺度不变性噪声估计方法得到随机噪声的噪声水平估计,并根据此估计值来归一化加权核范数最小化算法的保真项,继而对地震数据进行去噪处理。理论模型试验和实际数据应用结果表明,该方法能够根据噪声水平自适应地衰减地震数据中的随机噪声,并保持地震反射中的不连续性信息,实现对地震数据的盲去噪处理,为后期的构造解释、断层和断点识别、层位追踪、几何属性提取等提供良好的基础数据。
Random noise in seismic data has a negative effect on the analysis of effective signals,especially that of discontinuous signals.The scale-invariant noise estimation method can estimate the noise level in complex low-noise data.The weighted nuclear norm minimization method can highlight important information in the data by setting different weights,based on data differences acquired from singular values of the matrix.In this paper,a weighted nuclear norm minimization filtering method based on the estimation of the noise level is introduced.The noise level is estimated using the scale invariant noise estimation; subsequently,the noise level estimation is used to normalize the fidelity term of the weighted nuclear norm minimization filtering algorithm; finally,the seismic data are filtered.Numerical experiments and analyses of actual data showed that the method can adaptively attenuate random noise in seismic data according to the noise level.Moreover,it can maintain the discontinuous features of seismic reflection information,so as to achieve a blind denoising processing.The processed seismic data can provide data support for the interpretation of structural features,for the identification of faults and discontinuities,for layer tracing,and for geometric attribute extraction.
国家自然科学基金项目(41504097,41874153)资助。