iraf.net

Hi,
I'm using IRAF again after a long break, so there's probably a simple solution to this problem but I'm stumped:
I'm using imalign to register some dithered images. It finds a good alignment solution for each image, but I can't get it to perform the shift: the output images are trimmed but not shifted (i'm using shiftimages=yes, and imalign informs me that it's shifting the images). I also noticed that the output image files are nearly twice the size of the originals, but seem to have just one HDU corresponding to the unshifted images.

I'm running v2.16.1 on 32-bit Debian.

Thanks,
Nick

Could you upload some sample data files and the parameters you are using to the anonftp at ftp://iraf.noao.edu/pub so we can try to reproduce the problem? Also, how big are the expected shifts and what are you using to determine they haven't been shifted?

Sure - thanks. I've uploaded a file nick.zip containing some samples. I'm trying to align the three images image_{1|2|3}.fits. The file ref_sources contains coordinates of sources in the reference image (image_3), and the file estimated_shifts contains approximate shifts for each image with respect to the reference. I'm aligning them using:



This is some of the output I get:



The output images are image_{1|2|3}_aligned.fits. When I view them using gaia or other image software (gimp), the sources are all still at their original pixel coordinates. When I combine the images using imcombine, each source is duplicated because the images aren't aligned.

Cheers,
Nick

Just to follow up on this, in case anyone else has a similar problem:

I eventually figured out what was happening. The estimated shifts I gave had the wrong sign, so imcentroid was looking in the wrong direction for an alignment solution. When I changed the sign on the shifts the images were aligned correctly. It was confusing because even though the initial shift estimates were wrong, imcentroid didn't report a problem and converged on an incorrect solution that was close enough to the (wrong) initial shift estimate that it looked right. Anyway, all sorted now.