[argyllcms] Re: Argyll 1.5.1: artifacts in dark shadows - It got worse in 1.6!

  • From: János, Tóth F. <janos666@xxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 17 Oct 2013 14:48:05 +0200

I personally use an even higher resolution.

I modified the source code to increase the cLUT resolution from 45^3
to 66^3 for the Ultra setting. I also set the resolution of the in/out
curves from 2048 to 4096 and I set the curve order to 65 (cLUTres-1)
from ~25 (was it 25?). Oh, and I also set the gamut map resolution to

I ask targen to generate 66 gray single channel + 16 multidim steps
for display profiling. This seems to work fine for me. (It looks
acceptable even for profiling an uncalibrated laptop display).

I haven't figured out a better alligned setup for madvr calibration
but this also seems to work acceptable. Even though madvr uses TV
levels internally, the display is still 0-255, and since madvr use
high precision internally (and expands the 16-235 input to 0-255), I
guess it's fine to use the same 66^3 cLUT resolution. Or it's just
"enough by brute force", I don't know.

2013/10/10  <alberto.ferrante@xxxxxxxxxx>:
> Dear Graeme and all,
> yesterday I was able to do some further experiments and I have discovered
> that by using the -q u option in colprof almost all artifacts disappear. I
> just have some strange line on a very dark background, but it is much better
> than with -q h (-q m and -q l give reddish artifacts in the dark gray
> shadows). I tried to use -q u in colprof 1.4 and I got no artifacts and, in
> fact, the profile looks better than the one obtained by using -q h (I
> haven't had time to do any measurement though).
> Best,
>    Alberto
> Alberto Ferrante wrote:
>> Dear Graeme,
>> you are right: it was some months ago and I should have summarized the
>> history of this problem in my previous message, sorry! Let me do it now.
>> I will start by replying to your question: Since Argyll 1.5 I started
>> experiencing the problem of artifacts in the shadow areas. After some
>> tests we discovered that the problem was in the modifications to the
>> algorithm used to compute curves in the area near black that have been
>> introduced in colprof 1.5. Since we could not solve the problem (though,
>> the profile generated by using the "test" version of colprof 1.5.2, if I
>> remember well, was a bit better then the one generated by using 1.5.1),
>> I kept using all the up-to-date tools of Argyll except colprof. I was
>> successfully using colprof 1.4.x to generate my profile. I now see,
>> that, by mistake, I updated Argyll from the Fedora repository:
>> therefore, I wasn't using colprof 1.4 anymore, but 1.5.1: sorry, my
>> mistake! It may well be that by using colprof 1.4 I can still generate
>> good profiles even with the other tools from argyll 1.6, I will test
>> that! It remains the fact that the problem seems to be worse in 1.6
>> (i.e., when using colprof 1.6).
>> About the whole story: I sent to you test images, the profile, and the
>> data used to generate this profile. You could reproduce the problem,
>> but, unfortunately, you could not find a solution to it. Actually, I
>> have to say that I had similar problems (but not with shadows, with
>> bright red) on another monitor (the one of my laptop). My impression is
>> that (but I might be completely wrong!) sometimes targen generates
>> patches that are outside the gamut of the monitor (or, maybe, too close
>> to the limits), even though a reference profile is provided by using -c.
>> This causes the final profile to try exceeding the monitor capabilities
>> (and this, of course, also depends on colprof). Most of the times,
>> changing the algorithm for generating the patches improves things (on
>> the laptop I am now successfully using -r). Yesterday night I applied
>> the same strategy on my main monitor and, by using -q in targen, I
>> almost completely solved the problem. My current calibration flow is as
>> follows:
>> warm-up of the colorimeter:
>> targen -v -d3 -G -q -f5000 eizoCX240-riscalda
>> dispread -v -yb eizoCX240-riscalda
>> calibration:
>> dispcal -t6500 -gs -b100 -qh -Yp -v -e1 -yb eizoCX240
>> profiling:
>> targen -v -d3 -G -e8 -g32 -q -c eizoCX240.icc -f4096 eizoCX240
>> dispread -v -yb -Yp -k eizoCX240.cal eizoCX240
>> colprof -A "Eizo" -M "CX240" -D "$d - $s" -v -q h -d mt -aX eizoCX240
>> (eizoCX240.icc is the previous month profile)
>> B.t.w., thanks a lot for introducing the -Yp option!!
>> Best regards,
>>         Alberto
>>>> Hi all,
>>>> sorry to bring this old topic up again, but I was doing the
>>>> monthly re-calibration of my monitor and I noticed that, by using
>>>> Argyll 1.6, the problem with dark shadows got worse. I now obtain
>>>> artifacts even by using colprof from Argyll 1.4 (which was not the case
>>>> before) and the artifacts are much more evident. Artifacts now seem to
>>>> disappear (I have to check again more carefully when I get home
>>>> tonight, though) by using -al instead of -aX.
>>>> Is there anything I can
>>>> do to help solving this problem? Last time we could not find a solution
>>>> to it.
>>> Hi,
>>>         I'm not really following you. How can Argyll V1.4 be affected
>>> by changes in Argyll V1.6 ???
>>> I don't really recall where we got to last time, whether the problem
>>> was resolved, or whether I was unable to reproduce it.
> Tiscali Incontri by Meetic
> Offerta esclusiva pensata per te: 7 giorni gratis per fare nuovi incontri!
> http://k.ilius.net/?mtcmk=739565&fsid=080

Other related posts: