[argyllcms] Re: Color errors on < 100% stimulus
- From: János, Tóth F. <janos666@xxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sat, 11 Feb 2012 19:46:53 +0100
There is no optimal pattern size. Patter size should NOT matter at all.
When there is a difference, you should match the APL of the pattern
set with the APL of your normal content. But that's simply impossible
if you watch movies (or even a single movie all the time but with both
dark and bright scenes).
By the way, you are right. The color luminance control does not affect
the whole gradient, only roughly between 70 and 100% stimulus points.
I freshly calibrated a CG275W (I guess it's good enough to be a
reference monitor for this task) and a G30 with the same ColorMunki
spectro (I guess it's good enough...).
It's clearly visible how the G30 pushes the red with almost every
scenes but especially with low APL scenes. It is the most noticeable
on skin tones, especially with relatively dark scenes.
I paused the video where I saw a reddish face and tried to lower the
Red Luminance control on the G30 until it meets with the CG. I hit the
lower limit (-30 from the initial -5) before I could see any
difference. May be it was a little bit better but the difference was
minimal and it was still too far from the skin tone on the CG (which
felt more neutral without the red push). Then I set it to +30 and
again, almost nothing.
And it must be the luminance because I measured 17 points of the R, G,
B gradients and the chromaticity looked constant (I plotted the CIE
x,y coordinates on a graph and it was a fairly straight line). But
these measurements were taken with constant, small sized windows.
2012/2/11 Kristian Jörg <krjg@xxxxxxx>:
> You are right about that. What I ment was that I could adjust the gamma with
> the controls available and obtain a graph with mostly perfect gamma response
> - according to the patterns I use of course.
>
> I used APL patterns from the AVCHD disc which is what I have heard is the
> best for grayscale/gamma adjustments. For color window patterns are believed
> as the best. Although I am unsure of what size. Mayby you have an insight on
> this?
> There are many voices on what is the preferred patterns though...
>
> János, Tóth F. skrev 2012-02-11 15:31:
>
>> No, gamma is far from perfect.
>> Try to measure the grayscale with very small, normal and full screen
>> patches. You will get three different gamma curves and only standard
>> windows size will give you an approximately good enough pure-power
>> gamma function.
>> The same goes for the separate red, green, blue gradients. They all
>> depend on the pattern size and they never really follow the same tonal
>> response (only "close enough").
>>
>> 2012/2/10 Kristian Jörg<krjg@xxxxxxx>:
>>>
>>> Thanks for the answers regarding profiling. I will not profile the plasma
>>> as
>>> a PC monitor.
>>>
>>> About the VT30 I tend to agree. I just learned that there is a special
>>> firmware that improves color luminance on higher panel brightness. This
>>> patch is "only" for professional calibrators, and only for the US region.
>>> Although it is freely downloadable. I could not find a EU version of
>>> it... I
>>> also learned that there will not be further patching done as the VT50
>>> model
>>> is imminent. :( :(
>>>
>>> However I do feel that the controls are effective in what they should do.
>>> The problem is that the CMS controls one point. The rest of the intensity
>>> levels are not following suite...
>>>
>>> I ended up with a calibration that is no better than the builtin THX.
>>> Which
>>> in turn has the same color linearity errors as my own calibration.
>>> Sigh...
>>> VT30: Grayscale and gamma can be perfect. Colors - no.
>>>
>>> I hope I am wrong.
>>>
>>> János, Tóth F. skrev 2012-02-10 15:12:
>>>
>>>> I was mad about the VT30E, I said I would never see one of those again
>>>> after I tried to calibrate one.
>>>>
>>>> But interestingly, the G30E is not that bad. The peak error is under
>>>> dE2000=5, the average is acceptable.
>>>>
>>>> VT30 -> Way too much manual controls to achieve effectively nothing
>>>> or
>>>> make everything worse. And many manual controls are there to mask the
>>>> errors in the color calculations (like secondary color adjustments).
>>>> It's a dead end!
>>>>
>>>>
>>>> 2012/2/10 János, Tóth F.<janos666@xxxxxxxxxx>:
>>>>>
>>>>> XYZ and Lab profiles can do that.
>>>>> But you need to measure a lot of points and there is a big problem
>>>>> with plasma: Average Brightness Limiter. It rescales the gradient,
>>>>> causing tonal response and other color fluctuations. You can't correct
>>>>> for that.
>>>>> If you profile your display with small windows patterns, those will
>>>>> measure nicely through the profile evaluation but you will get high
>>>>> errors with bigger window sizes using the same profile.
>>>>> It would require a huge and power-hungry 4D LUT processing to correct
>>>>> for that (built from an insane amount of measurements).
>>>>>
>>>>> 2012/2/10 Kristian Jörg<krjg@xxxxxxx>:
>>>>>>
>>>>>> I have a question regarding monitor profiling versus calibrating
>>>>>> HomeTheater
>>>>>> displays.
>>>>>>
>>>>>> When I calibrate my Panasonic VT30 plasma (not connected to a PC) with
>>>>>> an
>>>>>> external program i.e ChromaPure I get dE errors in the range of 1.0 or
>>>>>> less.
>>>>>> The color calibration is done at 100% stimulus. But when I check the
>>>>>> color
>>>>>> balance at 75% stimulus I get errors of about 10 dE! Ouch! There is no
>>>>>> way
>>>>>> of correcting that in the TV set's controls even though it has a very
>>>>>> thorough CMS module.
>>>>>>
>>>>>> The question is when profiling a PC display with Argyll (or similar
>>>>>> profiling software), does it correct color at different levels of
>>>>>> stimulus
>>>>>> or only at 100%? I.e if I would us my TV set as a PC monitor would the
>>>>>> profile correct the color balance from 0-100% stimulus?
>>>>>>
>>>
>>>
>
>
>
Other related posts:
