Here, data regularization is performed by an operator called AMO or Azimuth Moveout Biondi et al. (1998). It also allows a local coherent stack that will reduce data volume. AMO, however, has a non-negligible computational cost in the whole imaging process (about 10% of migration cost). Effectively, introduced as the cascade of DMO and inverse DMO, AMO is a partial migration operator. With less accuracy, one can instead use a simple normalized binning procedure.
The processing scheme begins by creating a 5-axis
time-midpoint-offset grid () for the data
volume. Then, we apply a simple sequence NMO/AMO/
to
regularize the data on the grid. This gridding procedure concurrently allows
data resampling in common-midpoint and offset at the limit of
aliasing, thus reducing further the cube dimensions and lower
migration computational cost.
Marine data are usually concentrated within a narrow range of azimuth, as
opposed to land data. Here, besides regularizing data along space
axes, we use AMO to sum data coherently over the cross-line
offset axis hy; conventionally,
the subscripts x and y refer to the in-line and the cross-line
direction, respectively. Thus, we obtain 4-D common-azimuth data, for
which hy=0. After transformation to the frequency-wavenumber domain,
this 4-D common-azimuth regularized dataset
is the wavefield
recorded at depth z=0, to which CAM is applied.
Migration is then performed iteratively through common-azimuth downward-continuation of the wavefield Biondi and Palacharla (1996). The common-azimuth downward-continuation operator is derived from the stationary-phase approximation of the full 3-D prestack downward continuation operator. For more accuracy with lateral velocity variations, we use several reference velocities and interpolation as in the extended split-step method Stoffa et al. (1990). The following chart summarizes the preprocessing and imaging schemes:
![]() |
||