Abstract
Matching seismic wavefields lies at the heart of seismic processing whether one is adaptively subtracting multiples predictions or groundroll. In both cases, the predictions are matched to the actual to‐be‐separated wavefield components in the observed data. The success of these wavefield matching procedures depends on our ability to (i) control possible overfitting, which may lead to accidental removal of primary energy, (ii) handle data with nonunique dips, and (iii) apply wavefield separation after matching stably. In this paper, we show that the curvelet transform allows us to address these issues by imposing smoothness in phase space, by using their capability to handle conflicting dips, and by leveraging their ability to represent seismic data sparsely.
