Paper Summary

Reflection tomography is non-unique and a regularization term is usually added into its objective function as an additional constraint. The anisotropic Gaussian filter has been successfully employed as such a regularization operator. By orienting the smoothing axes along the local dip directions, the new Gaussian filter helps reflection tomography produce models that conform to the reflector structures. For efficiency purpose, the 3D Gaussian filter is factorized into three 1D filters in a non-orthogonal coordinate system. The dip oriented Gaussian filter also provides the freedom for applying spatially variable smoothing.