[Openicc] [Printing-architecture] [Gimp-print-devel] Colour

[Graeme Gill]

Chris Murphy wrote:

> So *if* there is a debate it's between 16bpc integer and 32bpc float, not 8bpc or
> 16bpc.

I disagree. 8 bits is perfectly adequate if the colorspace is close to
perceptual (ie. perceptually linearized device space), and can have
benefits in terms of file size, transmission speed and processing
speed, particularly for large or very large/high res. images.
(If file size really wasn't an issue, no-one would use JPEG,
and research into better image compression would have ceased).

Of course as soon as you convert out of such a space (ie. after
you've applied the calibration curves) you want higher precision,
but in many situations this can be done in one step during
screening.

Graeme Gill.