Structural interpretation - 3-D Seismic Interpretation - page 84

take place under close user control, so that difficult areas can be picked with continuous
control of the quality of the result.
The quality of an autotracked pick may be improved by pre-conditioning the seismic
data using image processing techniques. One approach, structure-oriented filtering, has
been introduced by Hoecker & Fehmers (2002) . The idea is to stabilise reflections in
the presence of noise, without smoothing over faults. The process consists of three
steps: analysis of the raw data to determine the local orientation of the reflectors, edge
detection to find reflection terminations, and smoothing of the data along the local
orientation without filtering across the edges detected in the previous step. As well as
removing noise, it is also possible to use such a filter to remove genuine but small-scale
features of the data, such as very small faults or small-scale stratigraphic features. This
opens up the possibility of an iterative approach to automated interpretation. In the
first pass, all the fine detail is removed, permitting a rapid autotracking of the main
horizons, and perhaps automatic fault tracking. This first result can then be fine-tuned
by repeating the process on a dataset with less aggressive smoothing, stabilising the
autotracking by using the result of the first pass as a seed grid. The process can be
repeated through several cycles of iteration, until either the data have been interpreted
to the required level of detail or the limit set by the noise in the dataset has been reached.

3.2.4 Attributes

A major advantage of workstation interpretation is that measurements of the seismic
loop being picked are simple to calculate and store. The most obvious is loop amplitude,
but loop width, average amplitude in a window below or above the horizon, and many
others are commonly available. The ability to see these measurements in map view
from densely sampled data is a key step in getting information from the seismic data
about porefill (presence and type of hydrocarbons) and reservoir quality (porosity,
net/gross, etc.). The way in which this can be done is the topic of chapter 5 . Amplitude
maps can also be the key to recognising stratigraphic features, e.g. channel systems. For
accurate work, it may be important to know how the software calculates loop amplitude.
Some early autotrackers would simply use the largest seismic amplitude seen at any
of the (usually 4 ms) samples within the loop; since there is unlikely to be a sample
exactly at the loop maximum, amplitudes were systematically underestimated. Modern
autotrackers fit a curve to the amplitudes at the samples within the loop in order to
estimate the true maximum value.
A different type of attribute is particularly important to structural mapping. It is
possible to analyse both the picked horizons and the seismic trace data themselves to
look for lateral discontinuities; we shall consider in this section those that are related to
recognising faults, but they can also be used to aid geological interpretation in general.
The simplest approach involves calculation for a picked horizon of the local dip
value and its azimuth (Dalley et al. , 1989) . Figure 3.19(a) is a sketch map of a faulted