|
|
|
|
Theory and practice of interpolation in the pyramid domain |
Now, we interpolate non-stationary data for two shots from one
synthetic and one field experiment. Being a frequency domain approach,
the interpolation in the pyramid domain forces us to decompose the data
in patches, or time windows, first. The size of
these windows is one second in time and 500 m in offset. Each window is
processed independently. Figure 6a shows the input
data for the synthetic example with a 50 m offset sampling. Its
amplitude
spectrum is displayed in Figure 6b:
some events are aliased for frequencies above 15 Hz. After
interpolation on a 25 m grid in Figure 6c, most of
the aliased energy is gone (Figure 6d),
while all the main events are preserved. Note that in principle,
more interpolation steps would be necessary to remove all aliased energy
(above 30 Hz).
Finally, we interpolate a shot gather from a field data experiment in
the Gulf of Mexico (Figure 7). The close-up in
Figure 7a shows primaries above 4 seconds and multiples
below. The
spectrum in Figure 7b shows some
aliasing for the slowest events around 40 Hz. We interpolate this shot
from a 26 m to a 13 m grid in Figure 7c. Most of
the aliasing artifacts have been attenuated (Figure
7d).
|
|
|
|
Theory and practice of interpolation in the pyramid domain |