Illiah, Thank you so much for the Color Checker Design Principles write-up!
I'm so curious about it.
BTW—Does anyone think it's weird that web image for the CC in that article
is an atrociously bad rendering of a Color Checker. I would like to think
that truly enlightened color mavens simple are no longer concerned with
earthly desires of of accurate rendering :) I jest! But it's a fact that
image is awful in several dimensions of laziness.
Roger, why are your working so hard on your camera renderings? Take a cue
from Munsell Labs and stop worrying about accuracy of rendered color! LOL!
Thanks again
/wire
On Tue, Jan 19, 2021 at 2:27 PM Iliah Borg <iliah.i.borg@xxxxxxxxx> wrote:
Dear Roger:
Your camera is very linear. When you increase exposure 2x, raw umbers are
increased by 2x, until clipping occurs.
I didn't say you need to jump into LUTs. I just said that Capture One may
offer limited functionality when used with matrix profiles.
Grey patches don't improve profile quality by much, so, yes, I referred to
18 non-neutral patches.
The design principles of the MacBeth colour chart can be found in the
original article published here
https://munsell.com/color-blog/color-reproduction-photography-printing-television-40-years-colorchecker/
- it deserves a very careful read, as it explains what were the colours
they thought matter most, how the other colours are introduced, what was
the target gamut, how the card was intended for film, etc.
Number of pigments: I've heard this from some folks very close to the
source, and to check it myself I did a regression from spectral reflectance
measured from different charts, the old one I still have (from 1977, kept
mostly in the light-tight envelope) suggests 6 pigments, the newer ones,
1997 and 2018, suggest 5 or 4 pigments in use (on those 18 patches).
Generally, when you are making a profile from an output file, you are
producing a correction profile, something to be applied to the output. Some
raw converters offer a mode to produce an instrumental output, in order to
give a user the capability to make an input profile.
Here is an Argyll-based GUI you may find useful for camera profiling
https://www.dpreview.com/forums/post/51765762 As it accepts CGATS files
from RawDigger (trial is OK), you can perform flat-fielding.
"a LUT-based profile from *more* color patches (>18) made using *more*
(>4-5) spectrally different pigments is better?" - you can get much lower
dE values with those if the card contains pigments or at least colours
present in the scene.
I prefer LUTs and matrices based on sensor spectral response, not on
shooting targets. Hence we've made this:
https://github.com/Alexey-Danilchenko/Spectron One of the benefits is
that now colour transform can be computed on the fly using light spectrum
measurements.
As to how Argyll processes a non-linear TIFF to produce a matrix profile
with gamma 1, Graeme is the authority; I simply put a question here hoping
he could answer. Depending on the internal workflow in a converter, you may
not even want the gamma 1 profile.
On Jan 19, 2021, at 3:55 PM, <graxx@xxxxxxxxxxxx> <graxx@xxxxxxxxxxxx>
wrote:
Dear Iliah:profile "route" (even if restricted).
Before jumping into LUT-base profiles, allow me to exhaust the matrix
(Incidentally, I have no idea how "linear" is the sensor on my camera)target" to test this hypothesis)
(I may waste my time but I sense I'd need to shoot some kind of "linear
different pigments is IMO not very useful.
You wrote:
A LUT-based profile from 18 colour patches made using 4 or 5 spectrally
wrong.
The above sentence needs some "decoding". Please correct me if I am
of the ColorChecker 24 chart -- i.e. the "chromatic patches"?
A -- "18 color patches", you must be referring to the first three rows
B -- "4 or 5 spectrally different pigments"?to expose this?
+ you have evidence to that effect? You have an analytical technique
+ you imply "A LUT-based profile from *more* color patches (>18) madeusing *more* (>4-5) spectrally different pigments is better?
gamma 1.
You wrote:
Linear response curve in Capture One doesn't mean that the TIFF is in
TIFF is NOT a Gamma = 1.0. Correct?
So, one source of "error" lies with CaptureOne "Linear Response curve"
This would imply that, given a TRUE gamma 1.0 TIFF, a Matrix profilewould better fit the data?
Which could imply that I need to search a RAW processor with true, gamma1.0 capabilities.
OR, if I want to stick with CaptureOne, I need to construct a profilewith a gamma curve + matrix or gamma curve + LUT, out of Argyll, as this
will be a better fit to the TIFF data?
Argyll.icm" -qm -as dsc_1545
This is colprof, by the way: -----------------------------------------
C:\Argyll\bin>colprof -v -D"NikonD810 Sun by Argyll" -O"NikonD810 Sun by
No of test patches = 242.879998
Find white & black points
Picked white patch 19 with dev = 0.61474740 0.61995770 0.59374070
XYZ = 0.81818420 0.85411140 0.67296580, Lab = 94.059997 -1.029995
Picked black patch 24 with dev = 0.06569350 0.06593931 0.06525877-0.939994
XYZ = 0.03250362 0.03362278 0.02896499, Lab = 21.440002 0.139987
Creating matrix...99.286429 -0.186930 -0.009200
100%
Matrix = 0.812593 0.375636 -0.247561
0.188279 1.354335 -0.594045
-0.048322 -0.181316 1.087328
Creating matrix and single gamma curve...
100%
Matrix = 1.375351 0.468084 0.108843
0.549794 1.773742 -0.335445
0.113546 -0.288550 1.978726
Gamma = 1.638076
Creating matrix and gamma curves...
100%
Matrix = 1.390687 0.471090 0.111372
0.557597 1.776575 -0.331841
0.113004 -0.290098 1.967767
Gamma = 1.652157 1.639000 1.630019
Creating matrix and curves...
100%
Matrix = 1.557083 0.495953 0.080219
0.616324 1.884987 -0.349716
0.092444 -0.315630 1.856321
Input offset = -0.025502 0.034841 0.011705
Output offset = 0.010943 0.002343 0.005195
0 gamma = 1.638627 1.725190 1.579940
1 harmonics = 0.301739 0.078588 -0.163133
2 harmonics = -0.150228 0.020485 0.585098
3 harmonics = 0.204813 0.211907 -0.070997
4 harmonics = -0.183206 -0.049656 -0.089493
5 harmonics = 0.023567 -0.179230 -0.024803
6 harmonics = 0.032559 0.051132 -0.015880
7 harmonics = -0.072549 0.107063 0.118868
Doing White point fine tune:
Before fine tune, rel WP = XYZ 0.94545055 0.98165887 0.80988814, Lab
After fine tune, rel WP = XYZ 0.96420288 1.00000000 0.82490540, Lab100.000000 0.000000 -0.000000
abs WP = XYZ 0.80226468 0.83844692 0.66071610, Lab93.383004 -1.202173 2.853609
Black point XYZ = 0.02930276 0.03022375 0.02551961, Lab = 20.1332820.286174 -0.485864
White point XYZ = 0.802265 0.838447 0.660716is around 1.65. Is that fair?
Black point XYZ = 0.029303 0.030224 0.025520
Done gamma/shaper and matrix creation
Profile done
Profile check complete, peak err = 8.409639, avg err = 2.301651
----------------------------------------------------------------------
As you can see above, according to colprof "estimates", my camera gamma
readers benefit from reading your comments 😊
Thank you in advance for your kind and patient help. I'm sure other list
On Behalf Of Iliah Borg
/ Roger
-----Original Message-----
From: argyllcms-bounce@xxxxxxxxxxxxx <argyllcms-bounce@xxxxxxxxxxxxx>
Sent: Tuesday, January 19, 2021 3:32 PMgamma 1. So, the question is: will Argyll compute a reasonable matrix
To: argyllcms@xxxxxxxxxxxxx
Subject: [argyllcms] Re: Digital Camera Profiling from ColorChecker 24
Dear Roger:
Linear response curve in Capture One doesn't mean that the TIFF is in
profile using a TIFF that is not linear?
the functionality using matrix profiles with gamma 1.8.
Capture One expects LUT-based camera profiles, and will provide most of
different pigments is IMO not very useful.
A LUT-based profile from 18 colour patches made using 4 or 5 spectrally
to decrease profiling error a lot. That's the first resource to improve
Accurate capture of the target (low flare, flat-fielding technique) help
profiles.
find :
Max dE2000 < 6 means a very good profile. 8 is acceptable.
On Jan 19, 2021, at 3:14 PM, <graxx@xxxxxxxxxxxx> wrote:
In Argyll’s excellent online documentation, with regards to colprof, we
1.0) curves. This may be useful in creating a profile for a device that is
“ The -am option will create a matrix profile with linear (i.e. gamma =
known to have a perfectly linear response, such as a camera in RAW mode.”
exported out of CaptureOne (with Linear Response curve).
I created a digital camera profile for my Nikon camera, based on a TIFF
DeltaE? Out of 24 patches?
I confess I’m rather “new” to camera profiling.
I was looking at the “statistics” line :
“Profile check complete, peak err = 8.409639, avg err = 2.301651“ So
this means 8.4 DeltaE as the largest error and an average of 2.3
DeltaE, so I manually took “reading” with the color picker in Photoshop,I wanted to convince myself that the largest error was indeed 8.4
having converted the image from the Argyll profile to Lab using AbsCol and
got :
ORANGE patch, a 7.53 DeltaE.
https://1drv.ms/u/s!AkD78CVR1NBqlrl0s9TCgXleuRqxsw?e=0aq60R
The peak error, as far as I can see through this method, is on the
work “better”?
Is there a way to improve the profile? Would another type of profile
/ Roger Breton
From: argyllcms-bounce@xxxxxxxxxxxxx <argyllcms-bounce@xxxxxxxxxxxxx>
On Behalf Of graxx@xxxxxxxxxxxx
Sent: Sunday, January 17, 2021 8:14 PM
To: argyllcms@xxxxxxxxxxxxx
Subject: [argyllcms] Re: Digital Camera Profiling from ColorChecker 24
Found the culprit 😊
https://1drv.ms/u/s!AkD78CVR1NBqlrlw81DJ6sZtI3dZ8Q?e=gkm3on
Much better.
I had the patches in the wrong order in the measurements file.
/ Roger
--
Best regards,
Iliah Borg
LibRaw, LLC
www.libraw.org
www.rawdigger.com
www.fastrawviewer.com
--
Best regards,
Iliah Borg
LibRaw, LLC
www.libraw.org
www.rawdigger.com
www.fastrawviewer.com
