Interpreters analyze migrated seismic data to map the subsurface geology. Unfortunately, the seismic data image, which is a back-propagated wave-field, does not represent all geological features equally well. The general wave-field character of the migrated image sometimes conceals important geological details, such as faults and river channels. Interpreters, however, seek faults and river channels since they indicate the location and properties of potential reservoirs.
Faults and river channels do not always cause easily discernible image patterns. Interpreters often delineate faults and river channels indirectly by identifying their effect on surrounding reflectors. The seismic wave-field often does not contain fault reflections, since the fault reflects seismic energy poorly or, if the fault dips steeply, the reflection energy is not recorded or imaged. A fault, however, offsets the neighboring geological strata, which causes a discontinuity of the seismic reflectors. Similarly, river channels usually cut their neighboring geological strata in characteristically meandering patterns. Interpreters search for these reflector discontinuities when locating faults or river channels in a seismic image volume. A seismic image volume that enhances the subsurfaces discontinuities enables interpreters to map faults and river channels quickly and accurately.
What computational process, then, enhances discontinuities? In general, a roughening filter, such as a gradient or a second derivative, enhances discontinuities. Moreover, an ideal roughening filter for detecting faults and river channels should
Discontinuities vary in size. Some faults are only a single sample wide while others span several pixels and appear as transition zones of rapid change. Ideally, a processor could tune the filter process to any target by setting a few well-understood parameters.
Claerbout suggests using a pair of PE Filters to enhance discontinuities in a seismic image volume. Data-adaptive PE filter techniques promise high spatial image resolution. In general, the PE filter output tends to have a white spectrum, but the filter response can be indirectly tuned by the choice of the filter shape.
In this article, we extend Claerbout's earlier work: we apply three filters simultaneously and we exclusively focus on image processing for geological interpretation. We resurrected Claerbout's earlier work because Bahorich and Farmer1995, Haskell1995, and Nissen1995 obtained impressive results from similar research. All the processes that attempt to enhance discontinuities are sensitive to noise. Marfurt1995 discusses difficulties with coherent noise caused by the recording geometry of the seismic data (acquisition footprint).