[argyllcms] Re: Conversion of spectral file to .cie
- From: Elle Stone <ellestone@xxxxxxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 14 Oct 2016 13:27:40 -0400
On 10/14/2016 11:52 AM, Ben Goren wrote:
On Oct 14, 2016, at 1:06 AM, Tomas Sobek <sankalpo@xxxxxxxxx> wrote:
My motivation is that there are supposedly slightly different variations of the
target and I thought if I use data that came with my one I will get hopefully
close to reality.
While there is measurable variation, it's well within the margin of error for
any graphic arts workflow. Indeed, you're all but guaranteed to have
significantly greater variation across the chart from uneven lighting and lens
imperfections. Much more significant is the fact that your lighting is
radically different from D50 (because D50 is an ideal which doesn't physically
exist), which itself requires special attention.
Creating good camera input profiles is difficult, and reflective charts are
extremely poorly suited to the task -- and 24-patch reflective charts are
basically useless for anything except photographing 24-patch reflective charts.
The only way to get not-miserable camera profiles is to use a workflow that
captures the camera's spectral response...and that's going to require either a
spectroscope or a monochromator, plus all sorts of other things.
(I should note that the ColorChecker Passport is a superlative standard
reference that plays an essential part in any color-critical workflow...but
that part is emphatically not in the creation of camera input profiles. Rather,
it's to normalize exposure and color channel balance, and possibly to fine-tune
scene illuminant characteristization.)
Without diving down a very deep rabbit hole, your best bet is to use Raw Photo
Processor with the profiles it ships with. Iliah Borg is making the best
general-purpose camera profiles of anybody these days. They're general-purpose
tools and come with all the limitations that implies; it's possible to craft
your own specific-purpose profiles that will outperform his...but doing so
requires a lot of equipment, knowledge, time, and effort -- and the payoff
isn't worth it unless you're doing something like fine art reproduction.
I have learned a lot from reading your posts to this list and I'm always
interested to read what you write - thanks much! for sharing over the
Deriving custom chart readings for the ColorChecker might not be not
worth doing for all the reasons you give. Also 24 patches is not a lot
of patches when making even a general purpose camera input profile,
though this forum has seen long debates over the adequacy of
ColorChecker's 24 patches vs other targets with more but perhaps lower
But I would like to ask - as your own knowledge deepens and as you
experiment with ways to make accurate profiles for your very controlled
studio lighting using your increasingly interesting (and complicated!)
laboratory setup - is it possible that you might be raising the bar too
high for the rest of us mere mortals who just want to make a general
purpose camera input profile that might be more useful and even perhaps
a bit more accurate than whatever input profile is supplied by whatever
raw processor we might be using?
Unless they make their own camera input profile, Linux users and people
using free/libre raw processors perforce use the camera matrices in
dcraw. I think ACR/Lightroom and a bunch of other proprietary raw
processors also use these matrices. Are you saying that there is
absolutely no point in trying to make a better general purpose input
profile than the matrices in dcraw? AFAIK, many (maybe even all?) of
these matrices were made by shooting a ColorChecker target.
I'm sure Iliah makes wonderful general purpose camera input profiles,
and if he wanted to make his profiles available for people who don't run
Mac, that might be a very nice thing. But right now advising people to
use RPP profiles seems equivalent to advising them to go buy a
MacIntosh, as RPP doesn't seem to be available for Linux or Windows.
Other related posts: