[fitsbits] CRPIX clarification

[William Pence]

In order to bring this discussion to a conclusion and finalize the new
FITS Standard document, I suggest 2 small modifications to the previous
proposal to clarify the use of the CRPIXj keyword.  These are in
response to the objections raised here by Steve Allen on 27-May-2008:

1.  Change the first sentence in section 8.1 from

   Rather than store world coordinates separately for each datum (e.g. as
   is done for random groups data), the regular lattice structure of a
   FITS image offers the possibility of defining rules for computing
   world coordinates at each point.

to

   Rather than store world coordinates separately for each datum, one
   may define rules for computing world coordinates at each point.

2.  Delete the phrase "in units of pixels" from the first sentence
in the definition of CRPIXj as shown here:

   CRPIXj - [floating-point, indexed, default: 0.0]
     Location of the reference point in the image for axis j
        \delete{, in units of pixels,}
    corresponding to $r_j$ in Eq. (8.1).


The full text of the proposed change is shown below. I don't think we
should get bogged down here in trying to expand this discussion to cover
other issues such as Fourier sampling theory and image display
conventions.  These issues go beyond the scope of what the IAU FWG
intended to be clarified when it conditionally approved the new FITS
standard last month.

Bill

=============================================================

Proposed changes to the wording of the new FITS Standard document
to clarify the use of the CRPIXj keyword.  The current version of this
document is available at http://fits.gsfc.nasa.gov/fits_draft.html.

1) Insert the following paragraph at the start of Sect. 8.1:

   Rather than store world coordinates separately for each datum, one
   may define rules for computing world coordinates at each point.
   As stated in Sect. 3.3.2 and
   depicted in Fig. 3.1, image array data are addressed via {\em integral
   array indices} that range in value from 1 to NAXISj on axis j.
   Recognizing that image data values may have an extent, for example an
   angular separation, spectral channel width or time span, and thus that
   it may make sense to interpolate between them, these integral array
   indices may be generalized to floating-point {\em pixel coordinates}.
   Integral pixel coordinate values coincide with the corresponding array
   indices, while fractional pixel coordinate values lie between array
   indices and thus imply interpolation.  Pixel coordinate values are
   defined at all points within the image lattice and outside it (except
   along {\em conventional} axes, see Sect. 8.5). They form the basis of
   the world coordinate formalism in FITS depicted schematically in Fig.
   8.1.

2) Remove the italics from "pixel coordinates" on p75 (they are used in
    the first mention above).

3) Change the definition of CRPIXj on p77 to:

   CRPIXj - [floating-point, indexed, default: 0.0]
     Location of the reference point in the image for axis j
     corresponding to $r_j$ in Eq. (8.1).  Note that the
     reference point may lie outside the image and that the first pixel
     in the image has pixel coordinates $(1.0, 1.0, \ldots)$.

-- 
____________________________________________________________________
Dr. William Pence                       William.Pence at nasa.gov
NASA/GSFC Code 662       HEASARC        +1-301-286-4599 (voice)
Greenbelt MD 20771                      +1-301-286-1684 (fax)