Robust seismic data denoising based on factor group-sparse regularization Available to Purchase

Highly accurate and efficient seismic noise suppression methods play an important role in geophysical exploration. Effective suppression of seismic random noise can be achieved by utilizing the inherent low-rank characteristics of seismic data. Based on the traditional multi-channel singular spectrum analysis method, we introduce factor group-sparse regularization to improve the accuracy of robust low-rank matrix approximation, thereby improving the suppression effect of sparse noise. In the process of minimizing the objective function, we use proximal alternating linearization for subspace optimization. In practice, it is difficult to guess an exact rank, resulting in a large amount of time for parametric testing of conventional low-rank methods. The proposed method can get rid of the dependence on the selection of sensitive rank, and achieve effective suppression of seismic sparse noise on synthetic data and field data.