Sie sind auf Seite 1von 15

Dip Filtering

1
Overview
• It is frequently useful to remove steeply dipping events from
seismic data.
• This can be done using trace-mixing (i.e. averaging adjacent
traces).
• The effectiveness of the filtering can be controlled by adjusting
the weighting of traces in the mix and/or the number of traces
averaged.
• Best trace-mix is judged visually.
• Better, but more expensive, dip filtering can be done using a
frequency domain approach (fk filtering)

2
Dip Filtering - Synthetic
Example

Before Filtering After Filtering


3
Dip Filtering
- Real Data
Example

4
Trace Mixing
0.0 s
Enlarged
0

0.05
0.05

areas in

0
next 2 slides
0.1 s
0.1

0.1

0.15
0.15

0.2 s

0.2
0.2

0.25
0.25

0.3
0.3

0.3 s
0.4

0.35
0.35
0.4

time (s)
time (s)

0 ms/trace & 1:1:1 Trace mix


10 ms/trace events 10 ms/trace event 5
attenuated
Trace Mixing Detail (See Next
Slide)
• Each output cell on right hand panel
contains average of three adjacent cells
from left hand panel.
• Highlighted cells show high amplitude
events.
• Note that trace mixing has preserved
horizontal event but attenuated dipping
event.
6
Trace Mixing Detail
Trace average
0.08997784 -0.03151028 0.04197542 0.02451832 0.00445432 0.03283108 0.0119643 0.02085961
Time

-0.21050505 -0.23014901 -0.15514516 -0.2417893 -0.18760995 -0.19891523 -0.208489 -0.19133988


-0.16458458 -0.01586981 -0.09322489 -0.08447757 -0.04988526 -0.09047286 -0.0648111 -0.07389957
0.50455115 0.54909826 0.44848991 0.55041944 0.48495042 0.50119696 0.51532741 0.49373048
1.1025449 0.91055006 0.99257023 0.99562738 0.93039594 1.00097501 0.96651243 0.97618398
0.81959905 0.73279026 0.87046433 0.74897621 0.84474202 0.8068696 0.78494081 0.81331757
-0.0178747 0.26150449 0.17248146 0.1528375 0.40447664 0.13993176 0.19537655 0.18407127
-0.2769012 -0.09603344 -0.29489236 -0.14617758 0.12921171 -0.22134325 -0.18019304 -0.22078064
0.4006079 -0.20788651 -0.10394325 -0.05939615 0.19980474 0.02721792 -0.12354398 -0.10729745
1.08405175 -0.09603344 0.23363524 0.0416404 0.34819709 0.40218534 0.06148628 0.13206535
0.87899136 0.26150449 0.09850487 0.01169609 0.34506448 0.41148528 0.12364785 0.08577543
0.1153034 0.73279026 -0.26768311 0.01169609 0.26528419 0.19886339 0.15466824 -0.10987665
-0.26953513 0.91055006 -0.13922735 0.0416404 0.22868365 0.1746954 0.26688555 -0.0836694
-0.05939615 0.54909826 0.54909826 -0.05939615 0.17570828 0.34829511 0.34829511 0.17570828
0.16499795 -0.01586981 1.03390761 -0.14617758 0.03968815 0.3902431 0.29805295 0.3520412
0.04923019 -0.23014901 0.77032436 0.1528375 -0.07234255 0.19220234 0.23639749 0.30281829
-0.11831906 -0.03151028 0.13148934 0.74897621 -0.04423972 -0.0063673 0.28147013 0.21504933
-0.04678396 0.14521088 -0.18445781 0.99562738 0.04364099 -0.02693809 0.31376097 0.25977272
0.0904154 0.04586829 -0.05807497 0.55041944 0.04251409 0.02626756 0.17702946 0.34961628

Unfiltered 1:1:1 Trace Mix


7
Trace Mixing Weights
• A 1:1:1 mix implies simple 3 trace averaging.
• Other weights are possible. For example, 1:2:1 implies (left +
2xcentre + right)/4. This is weighted towards the centre and so
is less extreme in its effects.
• More than 3 adjacent traces can be mixed. E.g. 1:2:3:2:1 would
be a 5-trace mix.
• Usual approach is to try a number of different mixes and then
visually choose best.

8
Trace Mix Tests (Real Data)

Raw 1:6:1 9
Trace Mix Tests (Real Data)

1:3:1 1:1:1 10
FK filtering
• FK filtering is a frequency-based method for
filtering dips.
• It is much more expensive than trace mixing
but is more precise in its effects (e.g. a
specified range of dips can be surgically
removed).

11
FK filtering
• An FK transform is simply a 2-dimensional
Fourier Transform (explained on next slide).
• Once transformed, signals with different dips
appear in different parts of the transformed
amplitude spectrum.
• Hence, multiplying by a window removes
specific dips (cf frequency filtering in which a
window removes specific frequencies).

12
FK Transforms & Dip Filters
x x
t f
Fourier
Transform
all columns

xt domain fx domain

Fourier
Transform
all rows

x k
t f Reject
Inverse Reject
FK transform
Pass region

xt domain fk domain
13
Quick Test
• Go back to the trace mixing
example. Which would you
choose and why?

Raw 1:6:1

1:3:1 1:1:1

14
Quick Test Answer
• Problem here is steeply
dipping noise shown on the
raw section.
• This is obscuring strong sub-
horizontal reflections.
• There is no “correct” answer in Raw 1:6:1
this case as it is a question of
judgement but I like the 1:1:1
mix.

1:3:1 1:1:1

15

Das könnte Ihnen auch gefallen