next up previous [pdf]

Next: Future Work Up: Almomin: Tomography Previous: Synthetic example

Conclusions and Discussions

We showed that using a global scheme for picking the correlation lags can better detect the correct events and is more robust to noise and multiple events than a trace-by-trace method. However, since the global picking is done on sparse spline points, the picking surface is too smooth, which causes the result to have less detail than that of the local method and converges at a lower rate. Nevertheless, the estimated velocity model did correctly approach the true velocity model, even though we started with a constant velocity.

Moreover, the accuracy and smoothness of the global method is strongly influenced by the number of spline points specified by the user. Therefore, accurately estimating the smoothness of the picked cross-correlation surface can help in determining a good spline geometry.