Adaptive Convolutional Filter for Seismic Noise Attenuation

Abstract: Seismic exploration is currently the most mature approach for studying subsurface structures, yet the presence of noise greatly restricts its imaging accuracy. Previous methods still face significant challenges: traditional computational methods are often computationally complex and their effectiveness is hard to guarantee; deep learning approaches rely heavily on datasets, and the complexity of network training makes them difficult to apply in practical field scenarios. In this paper, we proposed a method that has only 2464 learnable parameters, and its parameter constraints rely on priors rather than requiring training data. The three priors we proposed can effectively attenuate random noise while significantly reducing signal leakage, ensuring that the separated noise remains as independent as possible from the processed data. We validated our method on National Petroleum Reserve-Alaska Survey, and the results indicate that our proposed approach effectively enhances noise elimination and seismic data resolution.