Rob C. wrote:
> The thought is by doing a reduction to stars first, and then matching image
> x,y to image x,y with a scale of 1, it should be possible to do the
> co-adding automatically by shifting the images by the match solution.
> Although maybe this is what you already did...
Yup, I used the positions of the brightest N stars in each sub-image
to define the shifts required to register them. "N" was something
like 140 or so, and each sub-image was about 500 pixels on a side.
One of the benefits of using a sub-image is that it decreases the
amount of rotation within the field (an effect Arne mentioned).
Rob continued:
> Is there interest in software which attempts "automatic" co-adding of
> images? I'd be happy to have a go, since I've got some groundwork already
> in place. Probably after I've got some quality results from my pipeline :-)
Arne answered:
> In addition [to rotation], you
> want partial-pixel alignment, which means you have to decide
> what algorithm you want to use for interpolation. I'd
> recommend looking at how iraf does it (imshift is the
> appropriate routine). Personally, this is a lot of work
> and I'd rather use a standard image processing package to
> do such reductions than to roll my own.
Said Rob, sadly, :-)
> Fine words of wisdom. I guess I really need to re-install IRAF and attempt
> to hurdle the learning curve...
Well, I _could_ point out that I used XVista
http://a188-L009.rit.edu/tass/software/xvista/index.html
to find stars in each image, and measure their positions ...
and to shift each image sub-section by a fractional pixel (row, col) offset.
One _could_ use the "match" program
http://spiff.rit.edu/match/
to determine the optimal shift and rotation between two images,
as Rob did (though in doing the co-addition for Disk Set 19, I used
another program, which searches only for translation, not rotation and
scale change).
So, there _are_ tools outside of IRAF to do some of these tasks :-)
On the other hand, the documentation of IRAF is a heck of a lot better
than that of my code :-( Oh, and IRAF handles floating-point images,
too, which can be important in some situations (like this one).
Maybe I should just be quiet ...
Michael Richmond