Nonequispaced discrete curvelet transform for seismic data reconstruction

Physical constraints during seismic acquisitions lead to incomplete seismic datasets. Curvelet Reconstruction with Sparsity promoting Inversion (CRSI) is one of the most efficient interpolation method available to recover complete datasets from data with missing traces. The method uses in its definition the curvelet transform which is well suited to process seismic data. However, its main shortcoming is to not be able to provide an accurate result if the data are acquired at irregular positions. This come from the curvelet transform implementation which cannot handle this type of data. In this thesis the implementation of the curvelet transform is modified to offer the possibility to CRSI to give better representation of seismic data for high quality seismic imaging.