Seismic interpreters and processors, and occasionally exploration managers and even technology managers, have been known to ask in an exasperated tone, "Why are there so many seismic imaging methods?" The implication, of course, is that there are too many; maybe one is the preferred number. (Maybe developing all the others has been a wasted effort?) Indeed, I believe one is the preferred number, but after nearly a century of imaging, we still haven't gotten the actual number down to one. In fact, the preferred imaging method of 2012 is probably not the one we really want to end up with. To understand why there are so many, we need to take a look at how the technology developed over the years.
A lot of seismic imaging, or migration, was done before we had computers to do the imaging with. This meant that the earliest imaging methods had to be based on very simple physical principles. The simplest principle, embodied in the very first migrated images, was to estimate where in the Earth's subsurface a measured reflection event may have reflected from, given its source and receiver positions. Enough independent measurements from a reflecting surface allowed us to map the surface. Surprisingly, this principle is in common use today; it is the basis of one of today's most widely-used imaging methods, Kirchhoff migration. Although it is not the most advanced method, users understand what Kirchhoff migration does and can often interpret through the artifacts it tends to produce, which is not the case with other, more advanced, methods. A second principle, later known as beamforming, summed neighboring reflection events along various trajectories that correspond to angles of incidence at the Earth's surface. Mapping these beams into the Earth along the incidence directions provided an imaging method that eliminated the need to migrate all the individual traces. Necessity was the mother of invention.
The digital revolution took place in the 1960s, and seismic processing became serious business. We learned about digital signal processing and its implications for seismic traces that are discretely sampled in space and time. We also learned how to use computers to model the wave equation. Jon Claerbout was one of the first to do this, and he was the first to apply this modeling to seismic migration. However, Claerbout realized that modeling the "real" two-way acoustic wave equation would be prohibitively expensive on the computers of the day, so he developed an approximate wave equation that modeled wave propagation in a generally downward direction or a generally upward direction, but not both at once. This was the first wave-equation migration method, and others soon followed; the race was on to produce the fastest, most accurate imaging method possible. Unfortunately, the fastest was never the most accurate, so a wide spectrum of methods was developed – some that could image steep dips, others that were very efficient, and still others that were very accurate but perhaps could not image steep dips. And the rules kept changing. Faster computers made last year's method obsolete. Prestack migration changed the economics and the objectives of seismic imaging. Depth migration became more important, and 3D completely revolutionized seismic exploration.
That is the answer to the question "Why so many?" In the meantime, the job of seismic imaging changed. At first, it was used to find structural targets such as anticlines and salt flanks. Gradually, careful attention to the wave equation enabled seismic imaging to become a tool for seeing through amplitude and phase distortions present in unmigrated data, and therefore to delineate more subtle stratigraphic traps like pinch outs. From there it was a natural step to investigate migrated amplitudes to see whether the lithology is associated with gas or oil. In a different direction, imaging became a tool to estimate seismic velocity inside the Earth. Gradually, imaging became part of a loop (sometimes seemingly endless) where prestack migration is used to estimate velocity that, in turn, is used to migrate some more. In fact, velocity has become such a crucial part of the imaging process that sometimes the estimated velocity is more important than the migrated image, as when we use velocity to estimate pore pressure. Meanwhile, the demands of structural imaging intensified, taking us from "Where is the flank?" to "Where is the base?" to "Where are the hydrocarbons beneath the base?" These demands have led the industry to develop accurate migration methods that can image very steep dips while handling extreme lateral velocity variations. At the high end, where the success of the most expensive wells depends in part on the quality of the seismic image, the emphasis is now on accuracy more than efficiency. Use of the full two-way wave equation (reverse-time migration) is common; however, we must remember that the acoustic wave equation is still an approximation to the true behavior of seismic waves in the Earth.
Seismic imaging continues to evolve today to satisfy the evolving requirements of seismic data analysis. Sparse acquisition of data from some land areas produces such poor-quality, noisy data that conventional migration typically fails to produce satisfactory images; some efforts at specialized stacking techniques before migration have enabled much better imaging. Unconventional resources, such as shale gas and shale oil, promise to make new demands on all of seismic processing, including imaging. In particular, determining sub-wavelength features such as fracture fields from migrated data will require amplitude-preserving imaging to appear relative early in specialized workflows. Generally, the trend toward analysis of migrated amplitudes will result both in better final images and more detailed estimation of lithology in moderately complex geology. Finally, the realization that the Earth is a lossy elastic medium will force us to keep working on even our most advanced imaging methods.