[fitsbits] Question(s) regarding development of proprietaryFITS manipulation software. . .

On Tue 2007/09/18 21:17:31 -1000, Maren Purves wrote
in a message to: Mark Calabretta
and copied to: gberz3 ,

I just reread that. I can see that. You're talking about something
like Wien's law, or maybe Wien's law in false colors, or maybe perceived
colors. My understanding of the OP's question/interpretation is that

Maren,

Sure, you can create a pseudo-colour lookup table anyway you like for
a regular FITS image - it doesn't have a natural colour representation
so just invent one to suit your purposes. No problems there.

But the paper (which needs a lot more work) is really about how you
could store colour information in FITS (e.g. jpeg to fits). At one
level, this could just be pretty-picture stuff - uncalibrated RGB
or CMYK. However, at another level you could use a colourimetric
representation to store a scientifically accurate, true-colour image,
i.e. as a CIE 1931 standard observer would perceive it, something which
you can't do with jpeg (as far as I'm aware, but can with PostScript
and probably PDF).

This is interesting because true-colour photography of astronomical
objects is not at all simple, even using three-filter composites which
are hard to calibrate. However, it should be straightforward to
generate a true-colour image if you have measured visible spectra,
after all they should contain all of the colour information. Just
multiply the spectra by the CIE 1931 colour-matching functions to
get colourimetric RGB or XYZ-tristimulus values. These would then
typically be converted to CIELAB or CIELUV for storage in three FITS
image planes. (For display you might then need to convert back to
RGB or CMYK, thereby distorting the colour information to a degree
dependent on the gamut of the output device.)

(Check wikipedia for an explanation of these terms.)

Cheers, Mark