Re: a colorspace quandry
Re: a colorspace quandry
- Subject: Re: a colorspace quandry
- From: Roger Breton <email@hidden>
- Date: Mon, 16 Apr 2001 21:00:42 -0400
Neil,
>
Doug at this point scanner profiles are perceptual and unidirectional so
>
even if you choose a relative intent there won't be a relative table. The
>
colours seen by the CCD are so much broader than the space colours that you
>
would want to perceptually maintain the integrity of the original.
Scanner profiles are _not_ necessarily *perceptual*. I know ColorSynergy
Scanner profiles, for example, always write relative colorimetric LUT in the
ICC header even though, when you inspect the header with a profile inspector
type of utility, you see an A2B0 tag there and you get the impression
because there is a "0" in that A2Bx tag there that it's got to be
perceptual, right? Wrong.
Read section E10 of the ICC spec. The only "rendering" is mapping the media
white point of the scanner media (i.e. GS1 of the IT8 target) to Lab = (100,
0, 0). The A2B0 tag always does this. However, at run time, the CMM may be
instructed instead to use absolute colorimetric, which scales white back to
its original value (< 100).
In general, a rendering intent is used for "rendering", that is, when
converting the PCS into a form that may be viewed. This happens with
printers and monitors, but not cameras or scanners. The media white point
tag of the scanner profile may be used by the CMM to interpret the PCS as
either relative colorimetric or absolute colorimetric. Both cases are
derived at run time from the A2B0 scanner tag.
Neil, I hope this is useful.
Roger Breton
Quibec Institute of Graphic Communications