Applications of Deep Learning-based Deghosting in Marine Seismic Data Processing

SUMMARY

We demonstrate the potential of using Deep-Learning (DL) based de-ghosting methods in marine seismic data processing. We use a field-informed synthetic modelling strategy to create a diversified training data set, and account for acquisition uncertainties such as variable depth streamer profiles, wave action and variable surface reflectivity both in 2D and 3D applications.

We present an application of source- and receiver deghosting on a UHR variable-depth streamer configuration, a 3D UHR example, and a field data application.