In the field of seismic data processing, seismic denoising is essential to improve the quality of seismic data. Various deep-learning methods have been developed, showing promising performance for seismic denoising. However, most deep-learning methods are based on linear neurons, resulting in limited expressive ability for complex seismic signals. To enhance denoising capability using nonlinear neurons, we develop a supervised quadratic U-shaped network (QUnet) for seismic random noise attenuation. The quadratic neurons in QUnet are represented by a second-order polynomial of input data and weighted parameters. Numerical synthetic seismic data experiments indicate that our QUnet method achieves higher signal-to-noise ratio results and preserves the continuity of reflectors effectively. We also test the effectiveness of QUnet on prestack field data and poststack seismic data. The detailed structures and weak signals of seismic data are better preserved by our QUnet method.

You do not have access to this content, please speak to your institutional administrator if you feel you should have access.