iraf.net

[b:e48ce6ea01]From: [/b:e48ce6ea01]licia troisi <troisi@mporzio.astro.it ([email]troisi@mporzio.astro.it[/email])>
[b:e48ce6ea01]Date: [/b:e48ce6ea01]September 27, 2005 7:42:35 AM MST
[b:e48ce6ea01]To: [/b:e48ce6ea01]fvaldes@noao.edu ([email]fvaldes@noao.edu[/email])
[b:e48ce6ea01]Cc: [/b:e48ce6ea01]testa Vincenzo <testa@mporzio.astro.it ([email]testa@mporzio.astro.it[/email])>, de Santis Cristian <desantis@mporzio.astro.it ([email]desantis@mporzio.astro.it[/email])>
[b:e48ce6ea01]Subject: [/b:e48ce6ea01][b:e48ce6ea01]astrometry with MSCRED[/b:e48ce6ea01]
Dear Mr. Valdes,
I work for the LBC group of the Rome Astronomical
Observatory, and I would like to ask for your help for a problem we are having
with 'mscred'.
We are developing a pipeline for the reduction of images
from the LBC instrument and in particular we are now working on the
astrometry procedure. We would like to use the 'mscred' package to correct
distortion and apply astrometric corrections on some sets of LBC simulated
images, but we did not succeed, because, at the end of the procedure, the
routine 'mscimage' could not correct the relatively severe distortions at the corners.
In order to clarify the problem, the simulated images are produced in the
following way:
1) a sky map is produced by using the GSC2 catalog;
2) the simulator applies the optical distorsion as known from the optical
design; in particular, they are described with a 5-th order Chebyshev
polynomial, that gives out small residual with respect to the ideal distorsion
matrix (below 0.1 pixel).
3) images are generated from the distorted map by projecting it on the four
camera chips.
The simulator uses the tasks 'geomap' and 'geotran' within iraf to simulate
the distorsion. For sake of completeness, this is the  parameter set used by
geomap:
xxorder=5
yyorder=5
xyorder=5
yxorder=5
xxterm="full"
yxterm="full"
function="chebyshev"
The amount of distorsion generated by the distorsion map is of the order of
15-20 pixel at the corners, with a pixel size of 0.224". The shape is pin cushion.
Since the header is not touched when the distorsions are applied, the
astrometric info is the original undistorted one, as obtained from the sky map
of GSC2.
On the image sets, the procedure described in the 'mscred' manual is used,
i.e.:
- mscgetcatalog -using GSC2 as reference catalog
- msccmatch with the following choice of parameters (this is an example of all
  the tests we made):
(nsearch=                   15) Maximum number of positions to use in search
(search =                  20.) Translation search radius (arcsec)
(rsearch=                  10.) Rotation search radius (deg)
                                # Fine Centroiding
(cbox   =                   11) Centering box (pixels)
(maxshif=                   5.) Maximum centering shift to accept (arcsec)
(csig   =                  0.1) Maximum centering uncertainty to accept (arcsec)
(cfrac  =                  0.5) Minimum fraction of accepted centers
(listcoo=                  yes) List centered coordinates in verbose mode?
                                # WCS Fitting
(nfit   =                    4) Min for fit (>0) or max not found (<=0)
(rms    =                   2.) Maximum fit RMS to accept (arcsec)
(fitgeom=              general) Fitting geometry
(reject =                   3.) Fitting rejection limit (sigma)
(update =                  yes) Update coordinate systems?
(interac=                   no) Interactive?
(fit    =                   no) Interactive fitting?
(graphic=             stdgraph) Graphics device
(cursor =                     ) Graphics cursor
- mscimage ,with the following parameters:
                                # Output WCS parameters
(wcssour=                image) Output WCS source (image|parameters)
(referen=        LBCIMSIM.fits) Reference image
(ra     =                INDEF) RA of tangent point (hours)
(dec    =                INDEF) DEC of tangent point (degrees)
(scale  =                INDEF) Scale (arcsec/pixel)
(rotatio=                INDEF) Rotation of DEC from N to E (degrees)
                                # Resampling parmeters
(blank  =                   0.) Blank value
(interpo=               sinc17) Interpolant for data
(minterp=               linear) Interpolant for mask
(boundar=              reflect) Boundary extension
(constan=                   0.) Constant boundary extension value
(fluxcon=                   no) Preserve flux per unit area?
(ntrim  =                    0) Edge trim in each extension
(nxblock=                 1020) X dimension of working block size in pixels
(nyblock=                 2300) Y dimension of working block size in pixels
                                # Geometric mapping parameters
(interac=                   no) Fit mapping interactively?
(nx     =                   10) Number of x grid points
(ny     =                   23) Number of y grid points
(fitgeom=              general) Fitting geometry
(xxorder=                    5) Order of x fit in x
(xyorder=                    5) Order of x fit in y
(xxterms=                 full) X fit cross terms type
(yxorder=                    5) Order of y fit in x
(yyorder=                    5) Order of y fit in y
(yxterms=                 full) Y fit cross terms type
Please consider that the images are 2040x4600 and we tried different choices
for 'interpo' 'nx', 'ny' and so on.
At the end the output mosaic is not corrected. By plotting a subset of GSC2 or
USNO catalog on it, we can see that the distorsion have been only slightly
reduced, to an order of 4-5 arcsec.
We noticed that in the help page of 'mscimage' it is stated that:
   "Only  the  linear  components  of   the   input   image
    coordinate  system  are used.  In other words, the linear scales and
    rotation of the coordinate system at  the  tangent  point  are  used
    along  with  a  standard  tangent  plane  projection  for the output
    coordinate system.  The resampling will remove any higher distortion
    terms."
Hence, we would expect that resampling takes care of the higher order
distortions.
Our impression is that we are doing something wrong. Do you have any
suggestion or comment about our problem?
Might it be that, for example, the amount of
distortion is too large for 'mscred' to correct?
If you want, we can send you some of our simulated images for testing.
Thank you in advance for your attention.
Sincerely Yours,
Licia Troisi
[b:e48ce6ea01]From: [/b:e48ce6ea01]Frank Valdes <valdes@noao.edu ([email]valdes@noao.edu[/email])>
[b:e48ce6ea01]Date: [/b:e48ce6ea01]September 29, 2005 5:16:32 PM MST
[b:e48ce6ea01]To: [/b:e48ce6ea01]licia troisi <troisi@mporzio.astro.it ([email]troisi@mporzio.astro.it[/email])>
[b:e48ce6ea01]Cc: [/b:e48ce6ea01]testa Vincenzo <testa@mporzio.astro.it ([email]testa@mporzio.astro.it[/email])>, de Santis Cristian <desantis@mporzio.astro.it ([email]desantis@mporzio.astro.it[/email])>
[b:e48ce6ea01]Subject: [/b:e48ce6ea01][b:e48ce6ea01]Re: astrometry with MSCRED[/b:e48ce6ea01]
Hello,
Let me see if I understand.  You are making simulated images which have distortions
introduced.  You then want to use mscimage to see how well the distortions can be
removed.  This is reasonable.
The part that I did not see mentioned is about the world coordinate system (WCS).  The various
mscred tasks are designed to use a WCS description in the headers.  The concept is that
the distortion terms are relatively fixed and are stored in the headers.  The main sources of
observational error are assumed to be a telescope offset (corrected by adjusting the WCS
"tangent point") and a linear transformation which accounts for rotations and refraction.
Msccmatch is meant to automate determining and correcting these errors leaving the
higher order distortion terms unchanged.  Then mscimage uses the WCS, now assumed
to be accurate everywhere in the images, to resample the data to an undistorted celestial
grid.
So if you don't have a WCS which represents the distortions in the simulated mosaic see
http://iraf.noao.edu/projects/ccdmosaic/astrometry/astrom.html
In particular you should use mscfinder.msctpeak with a projection of tnx (after you have
an approximate, possibly just linear, WCS).  In order to do that your simulation should include
the GSC stars you used initially and you would have a simple text file list of the RA (in hours)
and Dec (in degrees).
An important thing that people often neglect is to verify things by the simple test of overlaying
star positions on the data.  This is done with msczero and the 'm' key and a list of coordinates.
If you can do this and see the stars marked correctly everywhere in the distorted image then
mscimage will also work correctly.  If you see the positions are off in the corners or some other
pattern then you will have errors in the resampled image.
If I misunderstood what you have done then let me know.  I would be glad to get some sample
data and look at it myself to verify whether or not what you have done should work correctly
with mscred and, in particular, mscimage.
Yours,
(Dr.) Frank Valdes
[b:e48ce6ea01]From: [/b:e48ce6ea01]licia troisi <troisi@mporzio.astro.it ([email]troisi@mporzio.astro.it[/email])>
[b:e48ce6ea01]Date: [/b:e48ce6ea01]October 6, 2005 7:41:06 AM MST
[b:e48ce6ea01]To: [/b:e48ce6ea01]Frank Valdes <valdes@noao.edu ([email]valdes@noao.edu[/email])>
[b:e48ce6ea01]Cc: [/b:e48ce6ea01]testa Vincenzo <testa@mporzio.astro.it ([email]testa@mporzio.astro.it[/email])>, de Santis Cristian <desantis@mporzio.astro.it ([email]desantis@mporzio.astro.it[/email])>
[b:e48ce6ea01]Subject: [/b:e48ce6ea01][b:e48ce6ea01]Re: astrometry with MSCRED[/b:e48ce6ea01]
  Dear Dr. Valdes,
  first of all thank you very much for your attention and your suggestions.
About the WCS, the simulated images have a WCS inserted that should be
compliant to the requirements of most packages. In any case, attached you can
find the header of the simulated images of a field.
As you probably know, LBC is composed of four chips, one of which is rotated
90 degrees with respect to the others.
The attached headers are for the four chips, where #4 is the upper, rotated,
one.
If you want to have a look at the simulated images directly, I can store them
for you in a web repository since we are without anonymous ftp service at the
moment.
I am now following the procedure indicated by you in your mail message,
to obtain from scratch a WCS that simulates the distorsions of the camera and
I'll let you know the results with the new WCS.
Apparently, in the previous attempt, after the 'msccmatch' routine we were not
able to resample correctly the images with 'mscimage', especially at the
corners, where the distortions are a bit more severe.
Thank you again for your support.
Best Regards,
Licia Troisi