iraf.net

Hello once again!This time my problem is with matching together images with XYXYMATCH. I'm pretty sure that I'm just doing something stupid and not inputting information correctly, because I know that XYXYMATCH is a powerful package. Here is my situation:I have four images that I want to match together automatically without enabling interactive mode. After using DAOFIND to create a star list for each image, I try and use XYXYMATCH to match the coo files to the reference image. Two of the images are rather close to one another (image1 and image2) and the other two are as well (image3 and image4), but there is a large distance between the pairs. XYXYMATCH can match an image to it's close neighbor very well, but it cannot make the jump to link across the big gap.Now I thought that XYXYMATCH works by taking an initial guess first and then using the tolerance or triangles algorithm to find the shift (I'm using triangles). However, XYXYMATCH doesn't seem to actually perform its initial shift. Even when I know the exact jump between the two images (110 pixels to the left and 110 pixels up for example), the program doesn't seem to work. I have used the refpoints file:
[code:1:85bf72421d]512,512
402,622[/code:1:85bf72421d]
or tried inputting the jump into the parameters
[code:1:85bf72421d]
(xin = -110) X origin of input list
(yin = 110) Y origin of input list[/code:1:85bf72421d]
Even when I try to fake out XYXYMATCH by putting in a giant initial jump between two really close images, I get a great match.
[code:1:85bf72421d]Input: image4.coo Reference: image3.coo Number of tie points: 0
Initial linear transformation
xref[tie] = 100. + 1. * x[tie] + 0. * y[tie]
yref[tie] = -100. + 0. * x[tie] + 1. * y[tie]
dx: 100.00 dy: -100.00 xmag: 1.000 ymag: 1.000 xrot: 0.0 yrot: 0.0Matched triangles transformation
xref[tie] = 5.855748 + 1.000838 * x[tie] + 4.30653E-4 * y[tie]
yref[tie] = -15.01828 + -3.5629E-4 * x[tie] + 1.000539 * y[tie]
dx: 5.86 dy: -15.02 xmag: 1.001 ymag: 1.001 xrot: 360.0 yrot: 360.089 reference coordinates matched
ecl>
[/code:1:85bf72421d]
XYXYMATCH seems to not be performing the initial jump at all, and merely drawing triangles from its star positions to the reference image's. This is the problem since I do not want to match each image to its partner, but rather all four to image1.What can I do to get XYXYMATCH to perform that initial transformation? I'm sure that its simply some dumb mistake that I'm making, where I'm not formatting my parameters correctly. Below are my settings for XYXYMATCH. I would be glad to send you the four images that I'm working with if you would like. Can someone please help me to get XYXYMATCH to work properly?[code:1:85bf72421d]
input = "image3.coo" The input lists
reference = "image1.coo" The reference lists
output = "image3.match" The output matched coordinate lists
tolerance = 2. The matching tolerance in pixels
(refpoints = "") Optional list of reference points
(xin = -110) X origin of input list
(yin = 110) Y origin of input list
(xmag = INDEF) X magnification required to match input to refe
(ymag = INDEF) Y magnification required to match input to refe
(xrotation = INDEF) X rotation required to match input to reference
(yrotation = INDEF) Y rotation required to match input to reference
(xref = INDEF) X origin of reference list
(yref = INDEF) Y origin of reference list
(xcolumn = 1) Input list column containing the x coordinate
(ycolumn = 2) Input list column containing the y coordinate
(xrcolumn = 1) Reference list column containing the x coordina
(yrcolumn = 2) Reference list column containing the y coordina
(separation = 9.) The minimum object separation
(matching = "triangles") The matching algorithm
(nmatch = 35) The maximum number of points for triangles algo
(ratio = 8.) The maximum ratio of longest to shortest side o
(nreject = 15) The maximum number of rejection iterations
(xformat = "%13.3f") The format of the output x coordinate
(yformat = "%13.3f") The format of the output y coordinate
(interactive = no) Interactive mode ?
(verbose = yes) Verbose mode ?
(icommands = "") The image display cursor
(mode = "ql")
[/code:1:85bf72421d]Thanks so much,
Mike

Hello,Acutally I have found that getting spatial matching programs to work well, especially xyxymatch, is tricky. There are various parameters that have to be tweaked and there is no one magic set of parameters that works in all or most cases (otherwise that would be the defaults). I can look at you case if you send me the position lists for a pair which have the large separation and don't seem to work.Frank Valdes
valdes@noao.edu

I have emailed you the coo files at the email address you listed!

I believe that I may have found a solution, although not the answer to my question about the initial linear transformation.When trying to get XYXMATCH to match my coo files, I was using as many stars as I could believing that the more data I gave XYXYMATCH the better. This means I could have as many as 150 stars in my coo file and their coordinates. What I failed to realize was that the nmatch parameter in XYXYMATCH was in fact only taking the first 30 or 40 entries anyways. So what does this mean?The close images worked because their first 40 entries had more in common than offset images, since the stars are ordered by their place in the image and not their intensity. By drastically upping the threshold in DAOFIND, I could limit the coo files to only contain the 15 or 20 brightest stars. These would be almost entirely the same stars in both images, regardless of image shift, meaning that XYXMATCH would have far less false triangles to try and draw. Now, instead, the triangles are between similar stars only.The results: XYXYMATCH works! I hope this helps anyone else reading this in an attempt to get XYXYMATCH working for them.

Hello!Thank you for your post, it helped me understanding XYXYMATCH.I also found out that another way of tricking XYXYMATCH to give the right answer is to use a high number in nmatch. This way it will use more stars to do the fit.But this solution it will be more time consuming... But it can be useful if you are using many images, and can't make a list of the bright stars easily.

Hi Michael,I'm sorry for the long delay in replying. I received your coordinate files. The first thing I did was make them into images using mkobjects. I displayed and blinked them to see the offset pattern. However, what I noticed is some sources are common but others are not. If I can believe the magnitudes these are not always the faintest sources. Below is a gif of the three artificial images produced from image1.coo, image2.coo, and image3.coo. Could you tell me if the way these fields look makes sense. By the way I registered the images to the source in the middle.Yours,
Frank Valdes[img:dac59b2f91]http://iraf.noao.edu/valdes/medford.gif[/img:dac59b2f91]