Next: Acknowledgments
Up: R. Clapp: Regularization
Previous: Results
Three different methods to regularize seismic
data with a wavenumber based AMO operator are described.
The adjoint implementation is the most efficient for
creating a Common Azimuth but inefficient for creating
multi-azimuth data. The adjoint approach also shows
noticeable amplitude dimming and brightening due
to acquisition geometry. Formulating
the regularization problem as an inverse problem and then approximating
the Hessian with a diagonal operator provides better results.
The cost and memory requirements are significantly
increased but multi-azimuth data comes for free.
Estimating the model with a conjugate gradient solver
produces the best results.
Amplitude artifacts are virtually eliminated and
the frequency content is noticeably improved.
Next: Acknowledgments
Up: R. Clapp: Regularization
Previous: Results
Stanford Exploration Project
5/3/2005