Estimates of seismic coherence of 3-D data sets have provided a radically n
ew way of delineating detailed structural and stratigraphic features. Covar
iance matrices provide the natural formalism to extend the original three-t
race crosscorrelation algorithm to larger analysis windows containing multi
ple traces, thus providing greater fidelity in low signal-to-noise environm
ents. By use of 3-D phase compensation using Radon transforms, we exploit a
dvances made in the high-resolution multiple signal classification (MUSIC)
algorithms, originally developed for the defense industry.
All three families of multitrace attributes (coherence, amplitude, and phas
e) are coupled through the underlying geology such that we obtain three fam
ilies of complimentary images of geologic features that result in lateral c
hanges in wave form. The phase attributes of dip/azimuth and curvature allo
w us to image areas that have undergone folding or draping that can not be
seen on coherence or amplitude images. The amplitude attributes allow us to
image oil/water contacts or other areas of amplitude variation that may no
t be seen on coherence or dip/azimuth images.
Coupled with coherence and the conventional seismic data, these new multitr
ace dip and amplitude data cubes can greatly accelerate the interpretation
of the major features of large 3-D data volumes. At the reservoir scale, th
ey will be of significant help in delineation of subtle internal variations
of lithology, porosity and diagenesis. In computer-assisted interpretation
, we strongly feel these new attributes will become the building blocks for
the application of modern texture analysis and segmentation algorithms to
the delineation of geologic features.