Interval velocity estimation from beam-stacked data - an improved method
, by Biondo Biondi

Low-wavenumber components of the velocity function can be tomographically estimated from
beam-stacked data. The velocity estimation method does not require data picking; rather it
maximizes beam stack energy at traveltimes and surface locations predicted by the velocity
model.
The gradient of the objective function with respect to the velocity model is explicitly
computed using ray tracing. The objective function is maximized by applying a conjugate
gradient algorithm and a quasi-Newton algorithm sequentially. The former converges to a
robust estimate of the low-wavenumber components of the velocity model; while the latter
resolves the finer components of the velocity model determined form the beam stacks.
The inversion method is successful in estimating the velocity function from synthetic data
in a horizontally-layered medium.