Implicit Images in FITS?

[Howard A. Bushouse]

In article <338A4AC7.1FF70E8 at astro.umd.edu>, Peter Teuben <teuben at astro.umd.edu> writes:
|> Wayne Landsman wrote:
|> > 
|> > Is anyone aware of any proposals to allow "implicit images" in FITS?
|> > For example, to store an image of constant intensity, one could specify
|> > the size of the image and the constant value in the header.   This
|> > information is sufficient to reconstruct the image, so that the explicit
|> > data portion of the FITS file need not be specified.
|> > 
|> > --Wayne Landsman                           landsman at mpb.gsfc.nasa.gov
|> 
|> 
|> No, but I have a hard time cooking up such a proposal. Once the
|> BITPIX, NAXIS, NAXIS1, NAXIS2 etc. have been set to define the
|> image, it expects data....So an alternate IAXIS, IAXIS1, .... would
|> have to be defined.
|> The next thing this proposal would want to do it to allow the
|> image be an analytical expression. Another example of something
|> that can easily run over an 80-character card length.
|> 
|> - peter

Wayne and Peter,

The STScI recently adopted (as an internal convention only) such a
scheme, but have not proposed it as a supported mode. The convention
sets BITPIX to whatever is required for the implicit image, but NAXIS=0
because there is no actual image data. We then use the additional keywords
NPIX1, NPIX2 to define the size of the implicit image, and the keyword
PIXVALE to set its' (constant) value. Our data I/O interfaces are in turn
setup to recognize such an implicit image and when reading one, automatically
constructs the image portion to return to a calling program.

-Howard