Mitigating the gradient artefacts of Migration Velocity Analysis by Gauss-Newton update

This paper shows how the artefacts present in the gradient of the cost function when performing a Migration Velocity Analysis can be strongly attenuated by using a second-order Gauss-Newton scheme. The artefacts on the velocity gradient are strongly attenuated when the total gradient is deconvolved by the approximate total Hessian. At each iteration, a least-squares migration is produced rather than a migration. The proposed algorithm is illustrated on the Marmousi dataset.