Bayesian ground-roll seperation by curvelet-domain sparsity promotion

The removal of coherent noise generated by surface waves in land based seismic is a prerequisite to imaging the subsurface. These surface waves, termed as ground roll, overlay important reflector information in both the t-x and f-k domains. Standard techniques of ground-roll removal commonly alter reflector information. We propose the use of the curvelet domain as a sparsifying transform in which to preform signal-separation techniques that preserves reflector information while increasing ground-roll removal. We look at how this method preforms on synthetic data for which we can build quantitative results and a real field data set.