Flexibly preconditioned extended least-squares migration in shot-record domain

We have developed a method for accelerating the convergence of iterative least-squares migration. The algorithm uses a pseudodifferential scaling (dip and spatially varying filter) preconditioner together with a variant of conjugate gradient (CG) iteration with iterate-dependent (flexible) preconditioning. The migration is formulated without the image stack, thus producing a shot-dependent image volume that retains offset information useful for velocity updating and amplitude variation with offset analysis. Numerical experiments indicate that flexible preconditioning with pseudodifferential scaling not only attains considerably smaller data misfit and gradient error for a given computational effort, but also produces higher resolution image volumes with more balanced amplitude and fewer artifacts than is achieved with a nonpreconditioned CG method.