CONCLUSIONS

Velocity analysis using a conjugate gradient method gives a better resolution than classical stacking along hyperbolas. Furthermore it enables us to easily get the original data from the velocity analysis panel. Modifications of basic operators using weighting etc. may be useful if we want to use only one iteration.

A comparison of the various ways of sampling in the velocity domain shows that even sampling in slowness squared should be chosen. It would be worth trying this comparison for matrices of sizes corresponding to the real gathers, but the amount of computer time required for this task discouraged me from doing so.

The minimum velocity for velocity analysis should not be chosen too low, because the data that is obtained from the velocity analysis panel might be spoiled by noise.