[argyllcms] Re: help with camera profile
- From: Ben Goren <ben@xxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Mon, 30 Jul 2012 06:27:56 -0700
On 2012-07-30, at 5:34 AM, Iliah Borg wrote:
> On Jul 30, 2012, at 8:25 AM, Markku Kolkka wrote:
>
>> Pentax DSLRs support DNG and the latest model (K-30) dropped the support for
>> the proprietary PEF file format.
>
> I do not think DCP profiles are very useful because they tend to change the
> colour with exposure adjustments.
This has been my experience, too. And I've only been using DNG profiles as the
first step in the workflow because ACR has, until recently, been the
least-worst RAW converter I've used.
It seems to me that the problem that DNG is supposed to solve (the whole
brouhaha over ``scene referred'' v ``output referred'') is nothing more than
the obvious result of using the worng color model. And it's equally obvious
that DNG is the worng solution and will always have serious, inescapable, fatal
flaws.
RGB isn't a color model; it's a representation of either the signal gain sent
to the CRT or the signal gain read from the image sensor. And there's really no
more reason why an end user should even be aware of those numbers than we're
aware of the voltage thresholds corresponding to 1s and 0s in RAM.
Add an ICC profile to RGB, and you logically transform it into a subset of Lab
(I'm grossly oversimplifying), but you still retain the confusion of the
meaningless device-specific voltage readouts.
Lab is a great color space for print output, what with 0 = light trap and 100 =
PTFE -- and those values remain the same whether you're lighting up the print
with multiple-kilowatt high-intensity light or trying to read it at the bottom
of a coal mine.
But, of course, the actual light levels and the corresponding stimulation of
your retina is quite different between those two scenes, and Lab isn't even
close to a good color space to record what you see.
Recall that RGB+ICC is a screwed-up version of Lab, and now realize that that's
the output we're getting from our cameras after Bayer interpolation, etc.
What I want to know is why this should be so.
We already have a color model perfectly suited to recording such scenes -- a
number of them, with XYZ being the one I see the most. What's more, since the
camera is also recording exposure information (aperture, shutter, and ISO
setting), there's absolutely no reason why RAW converters shouldn't be
outputting straight to XYZ.
And, then, suddenly, all our problems with gamut mapping and white balance and
all the rest...they instantly vanish. All those calculations are already done
today by first converting tagged RGB into XYZ (again, I'm oversimplifying) --
but XYZ of a print lit with a standard illuminant, not the XYZ values of the
original scene...so I'm left wondering why we ever even bother with RGB in the
first place.
Sorry for the rant...I'll shut up, now...just don't get me started on people
who specifically request CMYK JPEG files....
Cheers,
b&
Other related posts:
