Simultaneous source separation using a new multichannel matching pursuit method with directional features

Abstract:

Simultaneous source technology allows us to acquire seismic data in a much more efficient way and saves considerable acquisition time. However, it is necessary to separate these data into their conventionally acquired equivalent state. Luckily, we can simply treat the deblending problem as a noise attenuation problem because the blending noise has been performed to be incoherent in some domains such as common receiver, common offset domains. Multichannel matching pursuit (MCMP) is a lateral coherency based technique and has been widely used in a variety of seismic applications such as seismic trace decomposition and denoising seismic records. It decomposes the signals into a series of wavelets namely atoms, but the atom is just the best match to the average of multiple traces with the same scale factor, translation factor, frequency factor and phase parameter at each iteration, which is not in accord with the real seismic records. In this paper, we propose a new multichannel matching pursuit (MCMP) algorithm with directional features for simultaneous source separation in common receiver gathers. The new MCMP uses local lateral coherence as a constraint and utilizes the maximum semblance coefficient within a multidirectional window as the best direction at each iteration. To verify the effectiveness of this method, we use Ricker wavelet to synthetize a simultaneous source data set and sort the data to common receiver gathers. Comparing the deblending results with multidirectional vector median filter (MDVMF) method, the new MCMP preserves more useful seismic signals, but some individual useful signals are not reconstructed successfully probably because of the zero padding influence. The real data examples also prove that the new MCMP is effective in practice for deblending.