Non-linear regularization in seismic imaging

Two complementary solution strategies to the least-squares imaging problem with sparseness & continuity continuity constraints are proposed. The applied formalism explores the sparseness of curvelets coefficients of the reflectivity and their invariance under the demigration-migration operator. We achieve the solution by jointly minimizing a weighted l1-norm on the curvelet coefficients and an anisotropic difussion or total variation norm on the imaged reflectivity model. The l1-norm exploits the sparsenss of the reflectivity in the curvelet domain whereas the anisotropic norm enhances the continuity along the reflections while removing artifacts residing in between reflectors. While the two optimization methods (convex versus non-convex) share the same type of regularization, they differ in flexibility how to handle additional constraints on the coefficients of the imaged reflectivity and in computational expense. A brief sketch of the theory is provided along with a number of synthetic examples.

https://slim.gatech.edu/Publications/Public/Conferences/CSEG/2005/Herrma...