An interpreter wants a 3-D image of the subsurface in which planar features (faults, salt dome flanks) are sharpened and coherent, while 3-D plane layer volumes are removed. My filter approach, however, removes all seismic events despite the fact that it is designed to remove only plane layers and to protect temporal coherency. Consequently, I failed to control the output spectrum sufficiently and the final images are so white that they cannot be used for geological interpretation.
As expected, temporal pre-whitening helped to preserve temporal coherency. However, the limited spatial coherency of planar seismic features enabled my PE filter to remove all features of interest.
The approach could possibly be enhanced by spatial pre-whitening or other methods of spectral spatial shaping (e.g. introduction of an additional filter gap). Unfortunately, such a spatial whitening further increases the number of parameters and complicates the pre-whitening filter step tremendously.
The number of interdependent parameters already inhibits an exhaustive exploration of all possible options. Having tested various sets of parameter combinations, I failed to produce attractive coherency images. Since the parameters are not independent of each other in their effect on the outcome, I doubt that a more exhaustive search would succeed.
Controlling an images output spectrum through PE filters has proven to be difficult. Better theoretical understanding of PE filters might clarify if a combination with a moving average approach would allow more finely tuned control.
Finally, I am considering building a series of small synthetic data volumes. Such synthetic image volumes would allow me to test the coherency attribute using PE filters or alternative methods in a more controlled setting.