Noncausal $f$ -$x$ -$y$ regularized nonstationary prediction filtering for random noise attenuation on 3D seismic data

Conclusions

We have proposed a novel method for seismic noise attenuation using $f$ -$x$ -$y$ NRNA for 3D seismic data. f-x-y NRNA is the 3D extension of $f$ -$x$ NRNA. By using more information to predict the seismic signal, the $f$ -$x$ -$y$ NRNA improves the denoising result for 3D seismic data. The varying coefficients of the f-x-y NRNA are smooth along space coordinates for a given direction. The smoothness is controlled by shaping regularization, which has the key parameter: the smooth radius. The smooth radius can be selected by user according to the smoothness of assumed coefficients. This approach does not require breaking the input data into local windows along space axis, although it is conceptually analogous to sliding spatial windows with maximum overlap. Execution time of $f$ -$x$ -$y$ NRNA is reduced by iteration inversion and shaping regularization. Synthetic and field data examples both confirm that the proposed $f$ -$x$ -$y$ RNA approach can be significantly more effective in noise attenuation and consistency improvement than $f$ -$x$ RNA for 3D seismic data. Therefore, it may be useful in conjunction with interactive interpretation systems and auto-picking tools such as automatic event tracking.

Noncausal $f$ -$x$ -$y$ regularized nonstationary prediction filtering for random noise attenuation on 3D seismic data