主成分分析和字典学习联合地震数据去噪方法

Seismic data denoising method combining principal component analysis and dictionary learning

为能获得高信噪比的地震数据,笔者提出了一种基于K-SVD字典学习和主成分分析(PCA)相结合的主成分字典学习算法。与K-SVD算法对误差项直接采用奇异值分解来更新字典原子不同,笔者采用PCA算法分解误差项,并使用第一主成分作为字典原子的更新。通过对复杂模型合成地震记录与实际地震记录进行对比实验,得出该方法较K-SVD算法信噪比大约提高1~1.5 dB,能更好地保护有效信号。

To obtain seismic data with high signal-to-noise ratio,a principal component dictionary learning algorithm based on K-singular value decomposition(K-SVD)combined with principal component analysis(PCA)was proposed.Different from the K--SVD algorithm that updates dictionary atom directly using SVD,the PCA algorithm is used for the decomposition of the error terms,and the first principal component was used as the update of dictionary atom.Through the experiments comparing complex model synthetic seismograms with actual seismic records,it is concluded that the new method could increase the signal-to-noise ratio of about 1~1.5 dB compared with K-SVD algorithm,which could protect the effective signals.