频域奇异值分解(SVD)地震波场去噪

SVD (singular value decomposition) seismic wave field noise elimination in frequency domain

无规则干扰噪声的频谱分带较宽,不存在相干性;但在同一地区地震信号的主频带区间大致相同,尤其是叠后数据,也就是说各地震道有效频带的频谱存在较好的相干性。本文基于这个特点,通过傅里叶变换,在频率域进行SVD变换,并设计滤波因子,提取目标信号的奇异值进行重构频域信号,然后再进行反傅里叶变换,实现地震波场分离与去噪。与传统的带通滤波相比,该方法能彻底压制主频带干扰噪声,保护高、低频有效信号。实际地震数据处理表明,在不破坏原始地质信息的前提下,能较好地提高地震信号的信噪比。

Random noise has a wide frequency spectrum so that it does not have coherence, but the main frequency band for seismic signals in the same area are almost the same, especially for post-stack data, in other words, the frequency spectrum of the effective frequency band for seismic data has a better coherence than the random noise. Based on the conclusion above, SVD transform is conducted in frequency domain through Fourier Transform, the filtering factor was designed, and SVD values of the target function were extracted to reconstruct signals in frequency domain, then inverse Fourier Transform was conducted, and seismic wave field separation and noise elimination were realized. Compared with traditional band filtering the new method could completely suppress interference noise in main frequency band, protect effective signals in high and low frequency. The field seismic data processing results show that under the prerequisite in which the geological information is not damaged, the applying of the new method in seismic data processing could greatly raise the signal to noise ratio.