I constructed a crude interval velocity model that varied with depth only from the stacking velocities on the paper plot that came with the data tapes. Figure 4 shows the velocity model. I migrated all constant-offset sections with this velocity. The migrated constant-offset sections were binned into groups of three and stacked, reducing the data volume and improving signal/noise. The resulting sections were saved to be used for velocity analysis. Figure 5 shows the stack of the nearest three migrated constant-offset sections. The pull-up effect of the high-velocity dome on the bottom of salt reflector is still evident because the interval velocity model used for migration didn't ``know" about the salt.
Figure 6 shows four selected post migration common-midpoint gathers. Varying amounts of residual moveout are evident. The signal quality of the base-of-salt reflection is poor for the midpoint beneath the thickest part of the dome, but hopefully there will be enough velocity information to get a meaningful update from the inversion algorithm. Figure 7 shows the stack of migrated constant-offset sections. Despite the velocity errors evident in Figure 6, prominent events can still be identified. This illustrates one of the advantages of using an intermediate quantity like residual NMO+DMO stacking velocity (Etgen, 1989) over directly picking event locations. It would be difficult to pick robustly the images of reflectors on individual sections; stacking can improve the reliability of the velocity information in the data. This will be even more evident on the residual NMO+DMO semblance panels in Figure 8.