[fitswcs] Status of WCS negotiations

[Mark Calabretta]

On Mon 1998/07/20 12:55:49 -0400, Doug Mink wrote
in a message to: fitswcs at fits.cv.nrao.edu

>it back into the image header.  My current program, IMWCS in the
>WCSTools package fits CRPIXn, CRVALn, and a CDn_n matrix first and
>writes that to the header.  As an option, a polynomial, which replaces
>the CD matrix, can be fit to the offset from the reference pixel CRPIXn,
>CRVALn.  Usually, I fit six terms each for the x and y axes: a constant,
>x, y, x^2, y^2, and xy.  I have experimented with x^3 and y^3 terms,
>but that did not make a significant difference in the accuracy of the
>fit in any of the images I have tested.  For now, I have used the
>keyword COn_i for the ith polynomial term (in the list above) for
>the nth axis, and implemented only for two dimensions., the goal

As a hypothetical example, this is how you would be to translate your scheme
to G&C formalism:

   1) Indicate what you're up to by adding a code to the CTYPEn.  For example,
      set them to RA---CFA, DEC--CFA.

   2) Define and publish an algorithm which implements "CFA".

   3) Encode the parameters used by "CFA" in header cards of the form
      WCSiinnn (instead of COn_i). 

However, rather than define a fairly special-purpose algorithm it would be
better to define something which was more generally applicable so others could
use it.  I believe the translation of the DSS projection I described in
previous email would already satisfy your requirements.  It includes terms up
to x*(x^2 + y^2)^2 but you don't have to use them all.

Cheers, Mark