Multichannel band-controlled deconvolution based on a data-driven structural regularization

Abstract

Sparse deconvolution methods frequently invert for subsurface reflection impulses and adopt a trace-by-trace processing pattern. However, following this approach causes unreliability of the estimated reflectivity due to the nonuniqueness of the inverse problem, the poor spatial continuity of structures in the reconstructed reflectivity section, and the suppression on the reflection signals with small amplitudes. We have developed a structurally constrained multichannel band-controlled deconvolution (SC-MBCD) algorithm to alleviate these three issues. The algorithm inverts for a high-resolution seismogram rather than the full-band reflectivity series, thereby reducing the multiple solutions in the inversion and enhancing the reliability of processing results. We also exploited a structural constraint term to guarantee the spatial continuity of the structures, and we enhanced the relatively weak signals. The reflection structure characteristics, defined and extracted from the observed stacked seismic data, are the core of the structural regularization item. We solved the cost function of the SC-MBCD by the alternating direction method of multipliers algorithm. Synthetic model and field data examples demonstrate the rationality of SC-MBCD and confirmed that the algorithm can provide a better inversion result than the conventional sparse spike inversion in terms of retrieving weak reflection events and guaranteeing stratal continuities.

Seismic noise attenuation by signal reconstruction: an unsupervised machine learning approach

“…In seismic exploration, signals recorded by near‐surface sensors are usually contaminated by random noise, which causes poor performance in subsequent processing procedures, such as absorption compensation, seismic attribute analysis, and generating high‐resolution image (Xu et al ., 2015; Du et al ., 2018; Ma et al ., 2019, 2020; Wang et al ., 2019). Therefore, removing random noise from received noisy data is an essential procedure.…”

A novel high-resolution imaging method based on sparsity in the time domain and spectral fitting in the frequency domain

“…This should also be evident as a continuous trait. Many scholars have conducted extensive research in this area (Heimer and Cohen, 2009;Gholami and Sacchi, 2012;Gholami and Sacchi, 2013;Li et al, 2013;Yuan et al, 2016;Du et al, 2018;Ma et al, 2020). The focus of these studies has mainly been on the spatial continuity of seismic data.…”