Estimation of surface-free data by curvelet-domain matched filtering and sparse inversion

A recent robust multiple-elimination technique, based on the underlying principle that relates primary impulse response to total upgoing wavefield, tries to change the paradigm that sees surface-related multiples as noise that needs to be removed from the data prior to imaging. This technique, estimation of primaries by sparse inversion (EPSI), (van Groenestijn and Verschuur, 2009; Lin and Herrmann, 2009), proposes an inversion procedure during which the source function and surface- free impulse response are directly calculated from the upgoing wavefield using an alternating optimization procedure. EPSI hinges on a delicate interplay between surface-related multiples and pri- maries. Finite aperture and other imperfections may violate this relationship. In this thesis, we investigate how to make EPSI more robust by incorporating curvelet-domain matching in its formulation. Compared to surface-related multiple removal (SRME), where curvelet-domain matching was used successfully, incorporating this step has the additional advantage that matches multiples to multiples rather than predicated multiples to total data as in SRME.