Random noise attenuation by $f$ -$x$ empirical mode decomposition predictive filtering

Conclusions

We have proposed a new denoising method suitable for complex subsurface structures. We demonstrate that the number of dipping events will affect the denoising performance of $f-x$ predictive filtering. We also give the definition of an EMD based dip filter and ascribe the effectiveness of $f-x$ EMD to applying a high-cut EMD based dip filter to seismic profiles.

By using the AR model to predict the steeply dipping event, $f-x$ EMDPF can deal with complex seismic profiles that conventional $f-x$ EMD can't handle. By applying an EMD based adaptive dip filter in advance, $f-x$ EMDPF can preserve more useful energy as compared with conventional $f-x$ predictive filtering. $f-x$ EMDPF is actually a modification to both $f-x$ predictive filtering and $f-x$ EMD, so it maintains the benefits of being convenient, data driven, whilst combining the dip-selection property of EMD with the power of the AR model used in $f-x$ predictive filtering.

Although the incomplete EMD described in this paper can improve computational efficiency, a great deal of time is still required to process the data. Currently, this time requirement is the major drawback of the approach. In addition, continued research is required in order to find an efficient thresholding method in the $f-x$ domain in order to improve the preservation of useful signal.

Random noise attenuation by $f$ -$x$ empirical mode decomposition predictive filtering