Accurate Velocity Estimation Through the Curvelet Domain

We present significant improvements in pre-migration seismic velocity estimation through the use of Curvelet transform in prestack noise attenuation. Seismic data is decomposed using the Curvelet transform, which has the capability of separating events having differing frequency, dipping angle and location. Curvelet transforms decompose data as a weighted sum of “Curvelets”, where each Curvelet is localised in both the f-k and t-x domains, and each weight consists of both amplitude and phase. The data are processed in the Curvelet domain by manipulation of these weights. Noise suppression via the Curvelet transform on prestack gathers has contributed to significantly improved pre-migration velocity estimation.