Fast And Efficient Structure Oriented Preconditioning With On The Fly Dip Estimation

Summary

We introduce a novel computationally efficient model preconditioner, the guided filter, a multi-dimensional, structure-oriented, and edge-preserving filter, with a self-adjoint implementation which does not require apriori knowledge of the structural dip field. Model preconditioners are essential in geophysical ill-posed inverse problem to constrain solutions and improve convergence. We generalize the guided filter from two to higher dimensions to use it with 3D seismic images and models. We prove that this filter is self-adjoint and its computational cost is independent of the filter size.