| Re: Registration: still looking (dis-similar images, local , variation) [message #8619 is a reply to message #8532] |
Tue, 25 March 1997 00:00   |
Achim Hein
Messages: 42
Registered: February 1996
|
Member |
|
|
Morten Bro-Nielsen wrote:
>
> Hi,
>
> I couldn't find the original posting. Could someone please send it
> to me, please.
>
> I am just curious to see what the fuss is about - maybe my 2 cents
> could make a difference ?
>
Original:
I am still looking for software or algorithms to do a good job of
image registration (warping one of the images so equivalent features
overlay each other) for those cases where the two images are not
all that similar (different sensors or from much different points
of view), and/or where there is a LOT of elevation-induced small
scale local variation, due to different projection-to-ground or
layover geometries.
Nice, rapid user interfaces (e.g., grab a point with a mouse
and drag it to the desired position on the other image,
while watching the two registered images blink back-and-forth
between each other) would surely help. The ability to apply
smooth warping splines to many small areas is a must.
Any ideas?
This has surely got to be one of the oldest problems in remote
sensing, and good solutions MUST exist by now. I promise to
post useful responses to the news groups.
------------------------------------------------------------ --------
Mitchell R Grunes, grunes@imsy1.nrl.navy.mil. Opinions are mine alone.
Answer 1:
Perhaps you should offer a monetary reward. :-)
Cheers!
David
-----------------------------------------------------------
David Fanning, Ph.D.
Answer 2:
We are working on an automatic registration module too. We have modules
for detecting layover and shadow areas, for converting slant range
heights, distances and amplitudes to ground geometry, for precision
phase unwrapping and last but least a precision processing algorithm.
The simplest registration (airplane geometry) works like this:
-evaluate range displacements
-design a linear function in Range
-get a constant azimuth value
-registrated pixel=rg_constant+range_linear*pixel_number
-look for a threshold i.e. 0.1 pixel
evaluate it iterativ range by range
But think of David Fannings post:
"Perhaps you should offer a monetary reward."
That's the point - or are there any precise SAR-algorithms for free?
Regards
Achim
--
_______________________________________________
Dipl.-Ing. A. Hein
PB2 / ZESS - Uni-GH-Siegen
Paul-Bonatz Str. 9-11
57068 Siegen
Phone: 0271/740-3362
Fax: 0271/740-2336
Mail: Hein@nv.et-inf.uni-siegen.de
_______________________________________________
Please have a look at our Web-Sites:
http://www.nv.et-inf.uni-siegen.de/pb2/
_______________________________________________
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
