Fast methods for rank minimization with applications to seismic-data interpolation

Rank penalizing techniques are an important direction in seismic inverse problems, since they allow improved recovery by exploiting low-rank structure. A major downside of current state of the art techniques is their reliance on the SVD of seismic data structures, which can be prohibitively expensive. Fortunately, recent work allows us to circumvent this problem by working with matrix factorizations. We review a novel approach to rank penalization, and successfully apply it to the seismic interpolation problem by exploiting the low-rank structure of seismic data in the midpoint-offset domain. Experiments for the recovery of 2D monochromatic data matrices and seismic lines represented as 3D volumes support the feasibility and potential of the new approach.