[argyllcms] Re: Using a RGB-Camera as colorimeter
- From: Ben Goren <ben@xxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 3 Nov 2017 08:31:37 -0700
On Nov 3, 2017, at 1:43 AM, Happatz, Steffen <steffen.happatz@xxxxxxxxxxxxxx>
wrote:
I'm trying to find out how good a rgb camera performs as an colorimeter for
prints.
Pretty miserably.
Somebody with very solid color science chops could rig up a way to do it
anywhere from kinda okay (using a perfectly controlled photo reproduction
environment) to superlatively (by building a large-aperature spectroscope and
just using the camera as the device to record its output).
But, anybody doing so...is going to be using a spectrometer to calibrate the
system...and the spectrometers of choice are going to be more than enough for
printer profiling. In other words, it'd be an educational exercise, an
engineering challenge, a lark...but not a way to make printer profiles.
May I suggest? If the goal is good printer profiles, get your hands on a
ColorMunki and don't look back.
I thought if I have 604 color patches und i know the corresponding
L*a*b-values there should be no problem to map this from RGB to L*a*b.
So why does the transformation perform so badly in the mentioned areas?
Argyll isn't doing a simple this-point-to-that-point
plop-it-on-the-chart-and-be-done-with-it puzzle job. It's attempting to model
real-world printers. And, first, there's never a perfect match in such a round
trip in actual profiling exercises -- else there'd be no need for profiling!
That means that over-fitting the model to the data points results in bad
profiles.
Graeme can explain the math much better than I can, or you can look to the
source for yourself.
I'm a student and quite new and unexperienced to this field. I just want to
make sure i didn't do something wrong with the usage of ArgyllCMS. Also i'm
happy for every input i can get.
Welcome to the rabbit hole! Color science is one of those subjects that is
much, much deeper than those unfamiliar with it even begin to realize. The good
news is that it's potentially as student-friendly as Newtonian Mechanics. There
are some comprehensible fundamental principles from which most everything else
can be derived. With a bit of ingenuity, you can build all sorts of advanced
tools on a ramen-and-popcorn budget. The math isn't too bad, especially with a
modern spreadsheet to do the calculations for you.
Or, from the other end...if you're not looking at this from a color science
perspective but a "make good prints" perspective...get that ColorMunki and
religiously follow the "Scenarios" part of the documentation; that'll have you
up to speed in no time.
Cheers,
b&
Other related posts:
