Regularizing time tomography with steering filters

Robert G. Clapp and Biondo L. Biondi

bob@sep.stanford.edu, biondo@sep.stanford.edu

ABSTRACT

Standard depth tomography methods often do not converge, converge slowly, or converge to a model that is geologically unrealistic. By using dip based steering filters rather than standard isotropic regularization operators, convergence speed and final model quality are improved. By formulating the tomography operator in two-way vertical traveltime ($\tau$,$\xi$) rather than in depth (z,x) coordinates, we can uncouple velocity estimation from reflector position estimation. In this coordinate system we avoid many of the problems associated with the depth tomography problem, and converge quickly to a reasonable solution.