Registering Images
Hooloovoo wrote on May 30, 2006
Hi,
I need to align several *.fits frames from a single telescope and then stack them. The problem is that there are small rotations from frame to frame so I cannot use imalign. I attempted to use geomap but I ended up with empty images so I must have messed it up somewhere. Is there someone that can give me some directions on the best way to proceed with this alignment? I will need to perform this task on hundreds of images so I'm wondering whether there is a way to automate the registration without manually selecting the reference stars.
Thanks in advance.
I need to align several *.fits frames from a single telescope and then stack them. The problem is that there are small rotations from frame to frame so I cannot use imalign. I attempted to use geomap but I ended up with empty images so I must have messed it up somewhere. Is there someone that can give me some directions on the best way to proceed with this alignment? I will need to perform this task on hundreds of images so I'm wondering whether there is a way to automate the registration without manually selecting the reference stars.
Thanks in advance.
Mike Fitzpatrick wrote on May 30, 2006
We'll try the simple answer first: If the iimages have a correct WCS, see the WREGISTER task to align them. The WCSFIXER web app (search through the stories for more in this) can (sometimes) create/improve the WCS but isn't yet automated for many images. Hope this helps.
-Mike
-Mike
Hooloovoo wrote on May 30, 2006
Thanks for the reply.
Alas, there's no WCS information on the headers at this time. Our team is planning to implement it in the near future.
I have been reading Lisa A. Wells' Rectifying and Registering Images using IRAF guide and she suggests using geomap and register. If I'm not mistaken, geomap requires as input a list of x,y coordinates, which can be txdump-ed from daofind. This should be relatively trivial to automate if it was possible to run daofind on all images first to get the correct coordinate lists. The problem here seems to be that the objects on each image have to be marked in the same order such that each line corresponds to the same object on each of the images. Are there any straightforward ways to deal with this?
Regards,
H.
Alas, there's no WCS information on the headers at this time. Our team is planning to implement it in the near future.
I have been reading Lisa A. Wells' Rectifying and Registering Images using IRAF guide and she suggests using geomap and register. If I'm not mistaken, geomap requires as input a list of x,y coordinates, which can be txdump-ed from daofind. This should be relatively trivial to automate if it was possible to run daofind on all images first to get the correct coordinate lists. The problem here seems to be that the objects on each image have to be marked in the same order such that each line corresponds to the same object on each of the images. Are there any straightforward ways to deal with this?
Regards,
H.
Mike Fitzpatrick wrote on May 30, 2006
Have a look at STARFIND for creating object detection catalogs, and XYXYMATCH to match them up (i.e. will solve the rotation/shift between images). It should be possible to automate this in a script and get the information needed to adjust each image to a reference frame.
-Mike
-Mike
Hooloovoo wrote on May 30, 2006
Hi Mike,
Our survey is imaging a crowded field andstarfind detects hundreds of stars easily. The two test images I am trying to align have shifts of about 12 pixels in the x and 50 pixels in the y axis, while one is slightly rotated.
I run xyxymatch with the following parameters:
input = image1.coo.1,image2.coo.1
referenc= image1.coo.1
output = image1.coo.2,image2.coo.2
toleranc= 50. The matching tolerance in pixels
(separat= 10.) The minimum object separation
(matchin= triangles) The matching algorithm
(nmatch = 30) The maximum number of points for triangles algor
(ratio = 5.)
image1.coo.2 should be identical to image1.coo.1 since that is the reference frame itself, image2.coo.2 should contain the coordinates of matched stars on the two frames. I checked the coordinates of the matched stars but they do not correspond to the same objects on the two images so somewhere the matching algorithm seems to have failed (I presume that it's because the field is crowded and there are too many triangles).
I suppose I could input the object coordinates by clicking with the cursor on a small number of uncrowded stars but that wouldn't help much since I need to automate this.
Any suggestions would be welcome. If I can match the coordinates I can then use geomap and gregister to complete the alignment.
best regards,
H.
Our survey is imaging a crowded field andstarfind detects hundreds of stars easily. The two test images I am trying to align have shifts of about 12 pixels in the x and 50 pixels in the y axis, while one is slightly rotated.
I run xyxymatch with the following parameters:
input = image1.coo.1,image2.coo.1
referenc= image1.coo.1
output = image1.coo.2,image2.coo.2
toleranc= 50. The matching tolerance in pixels
(separat= 10.) The minimum object separation
(matchin= triangles) The matching algorithm
(nmatch = 30) The maximum number of points for triangles algor
(ratio = 5.)
image1.coo.2 should be identical to image1.coo.1 since that is the reference frame itself, image2.coo.2 should contain the coordinates of matched stars on the two frames. I checked the coordinates of the matched stars but they do not correspond to the same objects on the two images so somewhere the matching algorithm seems to have failed (I presume that it's because the field is crowded and there are too many triangles).
I suppose I could input the object coordinates by clicking with the cursor on a small number of uncrowded stars but that wouldn't help much since I need to automate this.
Any suggestions would be welcome. If I can match the coordinates I can then use geomap and gregister to complete the alignment.
best regards,
H.
Mike Fitzpatrick wrote on May 30, 2006
Crowded fields can indeed be a problem, but keep in mind you only need to use enough stars to get a solution. Tasks like TSELECT and TSORT in the TABLES package will let you use e.g. only the 100 brightest objects.
The real problem though I think is in your 'tolerance' parameter setting of 50 pixels: what this means is any star within a 50 pix radius is a possible match. Try restricting this a bit more so the final solution is more accurate. Hope this helps.
Cheers,
-Mike
The real problem though I think is in your 'tolerance' parameter setting of 50 pixels: what this means is any star within a 50 pix radius is a possible match. Try restricting this a bit more so the final solution is more accurate. Hope this helps.
Cheers,
-Mike
Hooloovoo wrote on May 30, 2006
Hi Mike,
I tried your suggestion and it works fine on the test files :) Thanks for all the help.
I tried your suggestion and it works fine on the test files :) Thanks for all the help.
Last post on May 30, 2006