Seismic Random Noise Attenuation via Noise Assisted-multivariate EMD Based MSSA

Abstract

Seismic random noise reduction is a critical task for seismic data processing. However, because seismic data is a representatively broadband signal, it is challenging to distinguish and attenuate random noise that present throughout the whole frequency range. Furthermore, it might be challenging to adjust the settings of denoising methods. To overcome the aforementioned problems, we suggest a multichannel filtering approach, i.e., noise-assisted multi-variate empirical mode decomposition (NA-MEMD) based multichannel singular spectrum analysis (MSSA). To filter noisy seismic data, we use NA-MEMD to decompose the noisy seismic data into a number of band-limited intrinsic mode functions (IMFs) with various center frequencies and bandwidths. Note that multiple channels of a particular IMF have the same dominant frequency aids in more reduction of random noise. Then, for separating random noise and maintaining valid signals, we apply MSSA to each IMF. The denoising outcome is then achieved by adding all the filtered IMFs. To demonstrate the validity and effectiveness of the suggested approach, we apply it to synthetic and 3D real data, and compare with traditional denoising techniques.