Source deghosting of coarsely sampled common-receiver data using a convolutional neural network

It is well known that source deghosting can best be applied to common-receiver gathers, whereas receiver common-shot records. The source-ghost wavefield observed in the domain contains imprint of subsurface, which complicates domain, particular when subsurface complex. Unfortunately, alternative, is, often coarsely sampled, this as well. To solve latter issue, we have trained a convolutional neural network apply domain. We subsample all shot records with and without receiver-ghost obtain training data. Due reciprocity, these data are representative set for coarse validate machine-learning approach on simulated field gives significant uplift compared conventional deghosting. field-data results confirm proposed remove from sampled gathers.