next up previous print clean
Next: Conclusions Up: Sava and Guitton: Multiple Previous: Examples

Discussion

Primaries and multiples can have shapes which are neither parabolic, nor hyperbolic in the data space. All multiple suppression strategies based on PRT, HRT or similar methods approximate the data moveout and may fail in complex areas. Furthermore, primaries and multiples often have comparable shapes which are hard to discriminate. Primaries and multiples have different shapes in the image space: primaries are mostly flat and multiples are non-flat. This allows, in principle, for robust signal/noise separation strategies.

For complex geology, the multiples are better attenuated if the propagation effects are taken into account. This is why the Delft approach performs so well Verschuur et al. (1992). For 3-D data, the latter can be rather difficult to use because interpolating the sources and receivers on a regular grid is very expensive. Alternatively, we can migrate the data and do the separation in the image space with conventional Radon transforms, as we demonstrate in this paper.

Potential pitfalls for the multiple suppression strategy in the image space include situations where our velocity model is far from the truth. We encounter the theoretical possibility that some multiples are flat and some primaries are not flat. However, even in such situations, we can still discriminate primaries from multiples, given enough separation in the Radon domain.

Finally, we would like to point out that the strategy described in this paper, which is based on Radon transforms, is not ideal. Other more sophisticated signal/noise separation methods, e.g., methods based on patterns Guitton et al. (2001); Haines et al. (2003), can better handle the non-parabolic shapes encountered in the image space thus producing better separation results.


next up previous print clean
Next: Conclusions Up: Sava and Guitton: Multiple Previous: Examples
Stanford Exploration Project
7/8/2003