Noniterative f-x-y streaming prediction filtering for random noise attenuation on seismic data

Conclusions

In this study, we introduced a fast approach to nonstationary prediction filter for random noise attenuation in the 3D $f$ -$x$ -$y$ domain. The proposed method employs a local similarity to constrain the autoregression equation for nonstationary prediction filter in the frequency-space domain, which belongs to the streaming prediction theory. Constrained conditions in the 3D frequency-space dimensions guarantee the accurate estimation of adaptive prediction filters and reasonable prediction of complex structures. Instead of using an iterative strategy, the new analytical solution in the frequency domain for the least-squares problem allows the proposed method to reduce computational complexity significantly. The matching snaky processing path further improves the signal recovery ability of the three-dimensional SPF. Although the $f$ -$x$ -$y$ SPF shows similar accuracy to the $f$ -$x$ -$y$ RNA, the proposed method allows us to better balance the target event protection, random noise suppression, and computational efficiency. Numerical examples using synthetic models and field data show that the $f$ -$x$ -$y$ SPF can effectively attenuate random noise and protect valid information in the nonstationary seismic data. The proposed method is superior in terms of its low computational cost even when analyzing large-scale seismic data.

Noniterative f-x-y streaming prediction filtering for random noise attenuation on seismic data