From: "milton_aupperle" <milton@outcastsoft.com>
Date: April 22, 2005 3:00:53 PM MDT
To: Astro_IIDC@yahoogroups.com
Subject: Alignment
Hi Folks;
After listening to feedback on the alignment issue in Astro IIDC beta
2.0 from Alan and other, I think I've come up with a better
methodology than previous.
Alignment of images is a "tricky" thing to say the least - especially
with atmospheric (turbulence, clouds, fog, birds, helicopters,
satellites etc.) and scope movement (poor tracking, ground movement,
wind movement etc.) mucking it all up.
When Astro IIDC does it's alignment, it calculates a centroid using
various methods for each frame. The centroid calculation is usally
good enough to get it down to +/- a few pixels, but that isn't good
enough. So we also do a pixel by pixel comparison against a small
32x32 (1024 ) pixel image in the "master" frame and the rest of the
frames within a 64x64 (4096) pixel pixel search area. That works out
to 134,217,728 pixel comparions for each image processed.
The problem that Alan and others has been having is caused by how the
area is selected for the 32x32 pixel block in the "master" image. The
old method would simply use the centroid as the starting point for the
32x32 pixel block. However that may or may not contain any real image
information , like a planetary edge, ring edge or a crater edge etc.
So the new method does a search on the first frame to locate an area
of maximum contrast and then uses that as a comparator from then on.
To see the difference check out "Alignment.jpg" in the new "Other"
folder which shows the new and old alignment methods. For objects with
definite edges (like craters, ring edges) it can do a much better job
- but for diffuse gas bags like Jupiter, it probably won't make much
difference.
HTH..
Milton Aupperle