image orientation on the screen

[Mark Calabretta]

On Mon 1997/01/20 16:29:02 PDT, Michael Newberry wrote
in a message to: fitsbits at fits.cv.nrao.edu

>You are right about using the signs of CRDELT1 and CRDELT2.  But you
>are talking about World Coordinate keywords whereas I'm talking about
>keywords that do not require an world coordinate calibration.  Most
>CCD users do not WCS calibrate their images---all I want is North
>almost up and East almost left for *display* purposes.

The default image orientation is independent of the coordinate system; that
is, the orientation is not determined by the sign of the CDELTn (e.g. you
could have secondary axis descriptions with conflicting signs).  While this
does not preclude a particular FITS viewer from transposing an image in order
to transpose the coordinate system, you couldn't expect all FITS viewers to
do the same thing, and in general I think it unreasonable to expect them to
intuit an orientation from curvilinear coordinate systems which need not be
parallel to the pixel axes, nor even orthogonal.  You were on safer ground
with your X/YREVERSE proposal.

The image orientation convention was added to the WCS proposal to formalize a
long-standing defacto FITS convention.  Strictly speaking it has nothing to
do with WCS (and is unrelated to the FORTRAN array storage convention).

Mark Calabretta
ATNF