From: Jon Claerbout
bill Goodway of Pan Canada Petroleum
used shot interpolation to overcome aliasing of multiples on offset axis 


From: Bill Harlan <harlan@pospsv.dnet.dupont.com>
To: matt@sep.stanford.edu
Apparently-To: matt@sep.stanford.edu
Subject: perl and pef
Date: Tue, 26 Mar 96 19:30:06 EST

To: matt@sep.Stanford.EDU

                                Tue Mar 26 18:24:40 CST 1996

Matt Schwab,

Thanks.

Well it looks like the new version 5 of perl does not even
use that directory anymore, but uses a different one "../perl5"
Never mind.  I don't want to screw up your distribution,
and you cannot assume I know what I am doing.

Greg Lazear recently used a multi-channel interpolation filter
that did a great job of interpolating beyond aliasing.
However, when he tried to use it as a coherency filter,
it was terrible because it was able to subtract almost everything
as coherent.  He'll need to degrade the quality of the
prediction, next.


                                Bill Harlan
 
                          Email: harlan@sep.stanford.edu (forwarded)
                          Conoco Inc. 
                          1000 S. Pine (4449 RW)
                          Ponca City, OK 74602-1267
                          phone: 405-767-6053; fax: 405-767-2887
                          Personal: P.O. Box 442; Ponca City, OK 74602-0442
                          http://sepwww.stanford.edu/oldsep/harlan/index.html

To: matt@sep.stanford.edu
Apparently-To: matt@sep.stanford.edu
Subject: 2D coherency tests
Date: Wed, 27 Mar 96 01:11:36 EST

To: matt@sep.Stanford.EDU

                                Wed Mar 27 00:03:43 CST 1996

Matt Schwab,

Greg used 2D Gulf Coast data.  I do think that something
useful ought to show up on 2D if it can work on 3D.
His main interest is in interpolation, so I haven't persuaded 
him to try much coherency filtering.  

Could we get your ground roll synthetic, or even better,
the code you used to create it?  (It would be better if we
could write out our own format.)  If he manages to deal with 
the aliasing, then I'll tell you how.  If he can't, then
who cares?  He is using a Spitz-like trick, but without
any Fourier transforms.  He has a few more tricks, but I'm
not sure whether they help.

                                Bill Harlan
 
                          Email: harlan@sep.stanford.edu (forwarded)
                          Conoco Inc. 
                          1000 S. Pine (4449 RW)
                          Ponca City, OK 74602-1267
                          phone: 405-767-6053; fax: 405-767-2887
                          Personal: P.O. Box 442; Ponca City, OK 74602-0442
                          http://sepwww.stanford.edu/oldsep/harlan/index.html


From: jon
To: ray@sep.stanford.edu, matt@sep.stanford.edu
Cc: dave@sep.stanford.edu
Subject: missing data off end
Date: Mon, 25 Jul 94 21:36:20 -0700

I noticed a simple kludge for handling missing data off-end
while at the same time using interior boundaries such as for patching.

Let m=missing and k=known and your data layout is this:

        			  --> x
         k k k k k k m m m m m m m
         k k k k k k m m m m m m m
      |  k k k k k k m m m m m m m
      |  k k k k k k m m m m m m m
    t |  k k k k k k m m m m m m m
      |  k k k k k k m m m m m m m
      V

The kludge is to set a border of m's to a known zero.
(actually a tiny nonzero since we code m=0)

         k k k k k k m m m m m m 0
         k k k k k k m m m m m m 0
         k k k k k k m m m m m m 0
         k k k k k k m m m m m m 0
         k k k k k k m m m m m m 0
         k k k k k k m m m m m m 0

If a2=3, maybe two levels of zeros would be theoretically better,
but likely nothing bad will happen with only one level.

This suggests the code I used on the 3-D passive data
in TDF p.78 Fig passfil should work dandy on extending the Mobil data.
We only need to add the above padding.

