结合主成分分析的四维块匹配协同滤波三维地震信号去噪

Three Dimensional Seismic Signal Denoising Based on Four-Dimensional Block Matching Cooperative Filtering Combined with Principle Component Analysis

四维块匹配协同滤波(BM4D)用于地震信号去噪时,虽然性能良好,但需要预知噪声标准差。针对上述问题,提出结合主成分分析(PCA)噪声估计的BM4D三维地震信号去噪算法。该算法首先用PCA对地震信号进行噪声估计,然后将估计结果用于BM4D去噪。对人工合成与实际三维地震信号的去噪实验结果表明,本文算法具有可行性,既能取得很好的去噪效果,又能规避BM4D对噪声水平预估值敏感的局限性。与5种噪声估计算法的对比实验表明,本文方法在噪声估计时间和精度方面均具有优势。

Four-dimensional block matching cooperative filtering （BM4D） has a good performance when it is used for seismic signal denoising. But it has to predict noise standard deviation. To overcome this issue , we present a three-dimensional seismic signal denoising algorithm based on BM4D combined with principal component analysis （PCA）. We first use PCA to estimate the noise standard deviation of the seismic signal, and then use the result of estimation for BM4D denoising. The experimental results of synthetic and actual 3D seismic signal denoising show that the proposed algorithm is feasible and can not only achieve the good denoising effect, but also avoid the sensitive limitations of noise level estimation. Compared with other five noise estimation algorithms, the experimental results show that the proposed algorithm has advantages in both noise estimation time and accuracy.