Aaron:
The following is taken from the GX documentation for GRIDAGC. It's a fairly
complete description of the method used (probably good enough for you to
duplicate the method in code, if you were so disposed).
If you have any questions on specific points, I'd be glad to help you. There
are no "limitations" to speak of,
but you can certainly do violence to your original grid with a judicious
selection of the filter size and maximum
gain correction.
I'm afraid someone else will have to help you with STAKPLOT, and I know
nothing of other methods
By the way, I got three copies of your message...
Now the documentation:
Input parameters:
Filter size Size of filter used to isolate signal from background.
Maximum Gain Correction Maximum gain correction to apply at any position.
Select Method Select method, "Full Amplitude" or "Local Amplitude".
The local gain is estimated using a moving window. (The window Size is
actually the closest odd number less than or equal to the input value so
that it may be centered over each position). In this window the average RMS
signal amplitude is calculated, and is taken to be an estimate of the local
signal gain. During the initial pass over the data, the maximum "gain" value
is recorded. In the second pass, all points are multiplied by the ratio of
the maximum gain to the local gain. A maximum gain correction is specified
to prevent the operation from "blowing up" in regions with little or no
signal.
Using Local Amplitude
In this case the method first isolates a "signal" component from a
"background" component. Inside the filter window, the best-fitting plane
(that which minimizes the RMS misfit with the data) is calculated. The
average RMS difference between the data values and the plane values in the
neighborhood is considered to be a measure of the local signal gain, while
the difference between the position's value and the best-fitting plane is
considered to be the signal itself.
The first pass over the grid determines the "signal" and the "signal gain"
at every position, and remembers the largest signal gain encountered. In the
second pass, the signal at each position is multiplied by the ratio of the
maximum signal gain to the local signal gain. The new "signal" is then added
back to the original background value to obtain the final value.
NOTE: Selecting Local Amplitude does NOT remove the background from the
final solution; just from the gain-estimation phase.
Choosing the Size is similar to choosing the corner frequency of a high-pass
filter; those features with wavelengths longer than Width are relatively
unaffected by the AGC, while features with wavelengths comparable to, or
smaller than Size will be most strongly amplified. The second method tends
to preserve more faithfully the longer-wavelength portions of the data.
Stephen Cheesman
(GRIDAGC author)
Geosoft
> -----Original Message-----
> From: Aaron Balasch [SMTP:[EMAIL PROTECTED]]
> Sent: Thursday, April 27, 2000 3:22 PM
> To: [EMAIL PROTECTED]
> Subject: [geonet]: Automatic Grid Compensation
>
> Can anyone out there give me information on AGC.(i.e. what is it supposed
> to do, how does it do it, what algorithm it uses, what are its
> limitations, what other algorithms are out there) or at least where to
> find information on AGC. Also, once I apply AGC to a grid using
> GRIDAGC.GX, is there any simple way, other than resorting to resampling
> the grid again, to get the matching stakplot of the grid?
>
> Aaron Balasch
_______________________________________________________
More mailing list info http://www.geosoft.com/support/listserv/index.html
