Conclusions

We have shown that non-stationary PEF's provide an effective means for suppressing multiples in the angle domain. Muting in the PRT domain provides models for multiples and for primaries that can be used as models for PEF estimation. Minor adjustments, such as lateral smoothing, before PEF estimation can greatly improve the final result. The results presented here show that the use of these PRT multiple and primary models in the pattern-based approach described here achieves better results than simple subtraction of the multiple model from the original data.

The purpose of a model for PEF estimation is to produce a PEF that captures the essential aspects of the signal or the noise. In many cases, this model need not exactly resemble the signal or the noise, but instead may be more effective if it is a simplified version of that which it targets. The lateral smoothing used here would not be an acceptable treatment of the actual data since it alters the frequency content and removes some amplitude information. It does, however, produce a model from which an effective signal PEF can be estimated. It is worth noting that this smoothing is akin to the stacking which is the final processing to be applied to each angle gather. This application to PEF estimation models, as part of the signal/noise separation approach described here, allows us to remove significant multiple energy prior to the final stack and produces an improved image. The cost of this smoothing, however, is a loss of some higher frequency energy.