Seismic Random Noise Attenuation Based on Streaming Prediction Filter in the F-X Domain

Guo LongYu,Liu Cai,Liu Yang,Zheng ZhiSheng,Wang QingHan

DOI: https://doi.org/10.6038/cjg2020n0030

2020-01-01

Abstract:Random noise is unavoidable in seismic prospecting. Conventional random noise suppression methods will damage nonstationary seismic signals with characteristics of time-space variation and finally affect accurate imaging of seismic data. Meanwhile, the current target of oil and gas exploration becomes "two widths and one high", with the increase of data scale, higher requirements are proposed on the processing efficiency of denoising methods. Therefore, there is a great significance to develop an efficient random noise suppression method for nonstationary seismic data. Prediction filters are widely used for seismic random noise attenuation. In this paper, we propose a new f-x streaming prediction filter based on the streaming framework. By establishing a prediction autoregressive equation in the frequency domain, the direct complex matrix inverse operation is used instead of the iterative method to calculate the nonstationary filter coefficients, and the time- and space-varying seismic events are predicted, computational efficiency of adaptive prediction filtering is significantly improved. By comparing with standard FXDECON method in industry and f-x regularized nonstationary autoregression (RNA) method, the results of synthetic and field data show that the f-x streaming prediction filter is able to better balance the relationship between effective signal protection, random noise suppression, and high efficiency calculation, the proposed method can get reasonable results.