Conclusions

Data regularization is an important problem when using migration methods that rely on the data being on a regular mesh. Traditional methods that apply the adjoint of a continuation operation such as AMO can lead to poor amplitude information in the regularized (and later migrated) cube. By setting up the regularization problem as inverse problem the amplitudes in the regular model space are significantly improved. The problem can be made computationally acceptable by intelligent parallelization and regularization choices.