next up previous print clean
Next: REFERENCES Up: Guitton and Claerbout: Galilee Previous: Approximating the model covariance

Conclusion

The interpolation of the Sea of Galilee dataset on a regular grid is addressed as a noise attenuation problem. We saw that the combination of preconditioning and Iteratively Reweighted Least Squares with the proper weighting function greatly reduces glitches and spikes. In addition, the introduction of a noise modeling operator that accounts for the inconsistency between depths measurements on different tracks greatly reduces the acquisition footprint. These two strategies help us to unravel meaningful geological features inside the lake and ancient shorelines. Our last attempt for improving the final image with a prediction-error filter as a preconditioner produces a more detailed map inside the lake but with smoother edges than with the Helix derivative. Unfortunately, our PEF estimation was not completely successful and we think that more work in that direction is desirable.

The lessons we learn from the processing of the Sea of Galilee dataset can be reused in our daily geophysical work. It teaches us that the residual should be always looked at to derive the correct weighting functions. It also seems to teach us that it is better to model and subtract the noise than to try to filter or weight it out of the residual.


next up previous print clean
Next: REFERENCES Up: Guitton and Claerbout: Galilee Previous: Approximating the model covariance
Stanford Exploration Project
7/8/2003