Conclusions and Future Work

Incorporating pseudo-primary data into a non-stationary prediction-error filter based interpolation method gives promising results for large gaps in the near offset. This problem would be very difficult to solve without the additional information provided by the pseudo-primaries, and the prediction-error filter approach eliminates a lot of the crosstalk that a simple cut-and-paste approach would have.T-x filters give a much nicer result than f-x, even with the greater dimensionality of the f-x filters. This is largely due to the issue of non-stationarity in time, which may be addressed by using the f-x approach in small time windows.

This method has been attempted on real data, and would initially appear to have the most benefit in the cross-line direction by creating pseudo-source lines where the receiver cables are, but the signal-to-noise ratio of initial attempts is very poor and not useful to show here. This would be equivalent to reducing the number of samples along the shot axis in Figure , which would clearly present problems. Also, the issues of cable feathering, swerving sail lines, 3D geometry and coherent noise all present problems with using this approach with real data.