yCMS - Color Management System [Archive] - Page 15

Fer

20th July 2011, 12:54

When I saw the comparison between the pure power and the BT.709 gamma curves with equivalent brightness, with the former looking better, I stopped asking you to remove it. However, I still think that you should give the true BT.709 transfer function as an option, because like it is now some people might wrongly assume that a pure power and a BT.709 with the same gamma values are almost the same, and we know that they aren't. ;)

Yes, I know that you don't want that many options inside madVR, and I agree with that on a user-friendliness level, but then some cases like this one from Fer might happen again.;)

Thank you yesgrey for the clarification. I rule out conf. 2) as an alternative.

jmone

20th July 2011, 12:58

Before measuring, you should always do the most basic adjustments with what the TV offers on controls, eg. Black/White level should be as good as you can get them.

Sadly, the NVIDIA option to change Limited to Full does not do what you think it would. It only affects EVR rendering output, not the actual output of the GPU. NVIDIA HDMI defaults to limited RGB, it only does Full Range if you create a custom resolution (or by some trick you have to apply to the driver inf file, i forgot what it was though). I always create me a 24p, 50p and 60p resolution anyway, so i don't suffer from that issue anymore.

Great :( .... FYI I've already had the TV Calibrated by an ISF Tech but I'm a complete nvidia newbie. With ATI GPU, I'd adjust my brightness and contrast using the test clips with it set to Full Range. I did try to do this on the nvidia (using madVR) but it did not seem to make any difference and in the end I had to leave it at Limited to be able to pick detail in dark areas. Does the uCMS only effect colour gamut or does it also adjust contrast / brightness?

Thanks
Nathan

madshi

20th July 2011, 13:04

I mean the image got brighter and washed out, with less contrast.

With conf. 2) I tried Gamma_Curve 1.0 2.08, where 2.08 is 2.6/1.25 (as I understand you suggest), and the result is different to the old conf. A). Again the observed gamma is lower than A). Even Gamma_Curve 1.0 2.20 gives a little bit lower observed gamma than A).

With conf. 1) the effect of the selection 2.1 gamma with a BT.609 curve in the madVR menu is what would be expected. The observed gamma is lower than with 2.6 gamma with a BT.609 curve in the madVR menu.
Ok, to clear this all up a bit, can you make screenshots of a video playing with:

(1) Your old config A).
(2) New config 2) with gamma processing enabled in madVR and set to a BT.709 curve of 2.6.
(3) New config 2) with gamma processing enabled in madVR and set to a BT.709 curve of 2.1.
(4) New config 2) with gamma processing enabled in madVR and set to a pure power curve of 2.6.

I think when we have those images to look at, things might get a bit clearer. Thanks!

(Of course it would help a lot if your screenshots used the same frame, to make things really comparable.)

So the yCMS configuration files of A) and 2) are not equivalent? yCMS generate with them equivalent 3dlut files or not? Is not allowed now the command Gamma_Curve 1.0 in madVR?
The 3dlut files are very different. E.g. the old one was YCbCr -> RGB, while the new one is RGB -> RGB. The whole approach is different. The Gamma_Curve command should not be used for newer madVR versions, anymore, because what Gamma_Curve did when using A) is now done by madVR's "gamma processing". The purpose of this change is that you can switch gamma values and curve types in madVR on the fly now, which of course is much more comfortable than creating various 3dlut files with different gamma curves/values.

Fer

20th July 2011, 13:20

madshi

20th July 2011, 13:39

Sorry yes, meant 1).

yesgrey

20th July 2011, 14:08

Does the uCMS only effect colour gamut or does it also adjust contrast / brightness?
I guess you meant yCMS.;)

You can use it to adjust contrast / brightness, but that would be sub-optimal. contrast / brightness should be adjusted on the display because that's the only way for you to use its full dynamic range, otherwise you will use only a small part of it.

Let's imagine your display's ON/OFF contrast ratio is 10,000:1. If you have your contrast / brightness controls bad adjusted you might end up with a lower contrast ratio. Of course yCMS could change the image so that you would be able to full all colors without any clipping, but that would be at the cost of part of the contrast ratio, and your black level would be much higher than it could be, or the display would look dimmer than it could.

The only controls that you should set-up before using yCMS are these two, exactly to allow you to use the full dynamic range of the display.

alph@

20th July 2011, 19:31

http://www.zimagez.com/miniature/gamma5.png

after corect by ycms the gamma curve is unbalanced,in madvr 'pure power curve at 2.20'

http://www.zimagez.com/miniature/gamut0.png (http://www.zimagez.com/zimage/gamut0.php)

the corect is good in green,But the Reds were reduced

http://www.zimagez.com/miniature/gamutgraph.png (http://www.zimagez.com/zimage/gamutgraph.php)

the red delta is 2.3 before corect and 5.6 after why?

Fer

20th July 2011, 23:46

Ok, to clear this all up a bit, can you make screenshots of a video playing with:

(1) Your old config A).
(2) New config 2) with gamma processing enabled in madVR and set to a BT.709 curve of 2.6.
(3) New config 2) with gamma processing enabled in madVR and set to a BT.709 curve of 2.1.
(4) New config 2) with gamma processing enabled in madVR and set to a pure power curve of 2.6.

(1) The old config A):

http://img27.imageshack.us/img27/8892/13181498.th.jpg (http://imageshack.us/photo/my-images/27/13181498.jpg/)

(2) New config 1) with gamma processing enabled in madVR and set to a BT.709 curve of 2.6:

http://img405.imageshack.us/img405/4099/bt2601.th.jpg (http://imageshack.us/photo/my-images/405/bt2601.jpg/)

(3) New config 1) with gamma processing enabled in madVR and set to a BT.709 curve of 2.1:

http://img52.imageshack.us/img52/2452/bt2101.th.jpg (http://imageshack.us/photo/my-images/52/bt2101.jpg/)

(4) New config 1) with gamma processing enabled in madVR and set to a pure power curve of 2.6:

http://img37.imageshack.us/img37/3313/pp2601.th.jpg (http://imageshack.us/photo/my-images/37/pp2601.jpg/)

(5) New config 1) with gamma processing enabled in madVR and set to a pure power curve of 2.2:

I have added this screenshot since the experimental data of the measured gamma for config A) adjust to a pure power curve with an average gamma value of about 2.20 (see the measurements of post #238 for a Gamma_Curve 1.0 2.55). So this screenshot should be similar to the screenshot (1). As you can see both are very similar. This fact seems to indicate that the implementation that madVR makes of pure power curves for gamma mangement is correct.

http://img155.imageshack.us/img155/9577/pp2201.th.jpg (http://imageshack.us/photo/my-images/155/pp2201.jpg/)

madshi

21st July 2011, 07:31

Thanks, those are interesting images. FWIW, if you set madVR to a pure power curve of 2.2, neither madVR nor yCMS are practically doing any gamma curve/value changes at all. Try disabling gamma processing in madVR. The image should look exactly like that with a pure power curve of 2.2.

After looking at those images I'm wondering whether yCMS or madVR do not handle the BT.709 gamma curve correctly. What do you think, yesgrey?

@Fer, could you do one more test: With your old config A), can you make a screenshot with Gamma_Curve 1.0 2.6 (wait, you already have) and another one with A) and Gamma_Curve 0.0 2.2? Please make sure you screenshot *exactly* the same frame. Then we can check if there's any difference at all. The screenshots of A) 1.0 2.6 looks pretty much identical to 1) 0.0 2.2 to me. So I'm kinda wondering whether A) 1.0 2.6 used a pure power curve of 2.2 for some weird reason?

Fer

21st July 2011, 07:59

@Fer, could you do one more test: With your old config A), can you make a screenshot with Gamma_Curve 1.0 2.6 (wait, you already have) and another one with A) and Gamma_Curve 0.0 2.2? Please make sure you screenshot *exactly* the same frame. Then we can check if there's any difference at all. The screenshots of A) 1.0 2.6 looks pretty much identical to 1) 0.0 2.2 to me. So I'm kinda wondering whether A) 1.0 2.6 used a pure power curve of 2.2 for some weird reason?

I will take the frame later, but although the static frames (1) and (5) are very similar, during movie playback a difference can be observed. Config (1) shows a little bit more depth and sharpness. So I think that (1) and (5) do not give exactly the same gamma curve. The difference must be subtle but it have an observable impact. Nevertheless the config of (5) may be also an interesting option for the movie playback.

:thanks:

Fer

21st July 2011, 18:17

@Fer, could you do one more test: With your old config A), can you make a screenshot with Gamma_Curve 1.0 2.6 (wait, you already have) and another one with A) and Gamma_Curve 0.0 2.2? Please make sure you screenshot *exactly* the same frame. Then we can check if there's any difference at all. The screenshots of A) 1.0 2.6 looks pretty much identical to 1) 0.0 2.2 to me. So I'm kinda wondering whether A) 1.0 2.6 used a pure power curve of 2.2 for some weird reason?

(6) The old config A) with Gamma_Curve 0.0 2.2 instead of Gamma_Curve 1.0 2.6:

http://img88.imageshack.us/img88/9393/a1pp220.th.jpg (http://imageshack.us/photo/my-images/88/a1pp220.jpg/)

The frame is similar to (1) and (5) frames of the previous post.

madshi

21st July 2011, 18:55

Unfortunately the old config A) with 2.6 is a different frame. Anyway, from what I can see, the following three images appear to have exactly the same gamma curve:

- old config 2.6 BT.709 curve
- old config 2.2 pure power curve
- new config 2.2 pure power curve

I can't see any difference in gamma between these images. And that means that IMHO you never really used a 2.6 BT.709 curve. You actually used a 2.2 pure power curve, without knowing it. At least that's my current impression...

@yesgrey?

yesgrey

21st July 2011, 19:48

the red delta is 2.3 before corect and 5.6 after why?
This is a recurrent question, so I think it's time for me to create a FAQ. I will try to do it soon, stay tuned.;)

yesgrey

21st July 2011, 20:21

After looking at those images I'm wondering whether yCMS or madVR do not handle the BT.709 gamma curve correctly. What do you think, yesgrey?
I think both are handling it correctly.

and another one with A) and Gamma_Curve 0.0 2.2?
@Fer,
Could you do one more with Gamma_Curve 0.0 2.1?
Thanks.

So I'm kinda wondering whether A) 1.0 2.6 used a pure power curve of 2.2 for some weird reason?
I think that didn't happen. Maybe on that frame a BT.709 2.6 looks similar to a pure power 2.2.

from what I can see, the following three images appear to have exactly the same gamma curve:

- old config 2.6 BT.709 curve
- old config 2.2 pure power curve
- new config 2.2 pure power curve

I can't see any difference in gamma between these images. And that means that IMHO you never really used a 2.6 BT.709 curve. You actually used a 2.2 pure power curve, without knowing it. At least that's my current impression...
Let's wait for the image I requested before taking any conclusions...

Fer

21st July 2011, 23:08

@Fer,
Could you do one more with Gamma_Curve 0.0 2.1?
Thanks.

Here is the screenshot.

(7) The old config A) with Gamma_Curve 0.0 2.1 instead of Gamma_Curve 1.0 2.6:

http://img829.imageshack.us/img829/9853/a1pp210.th.jpg (http://imageshack.us/photo/my-images/829/a1pp210.jpg/)

As could be expected the frame is different to (6) (Gamma_Curve 0.0 2.2). Now the image has lower contrast due to a lower gamma.

I think that didn't happen. Maybe on that frame a BT.709 2.6 looks similar to a pure power 2.2.

I agree with you. As I said before, although the static frames (1) and (5) are very similar, during movie playback a difference can be observed between both gamma curves.

jmone

22nd July 2011, 06:09

I'm stuggling (I still don't get it). I have my colormunki but just dont really know where to start (again if this is the wrong thread for a newbie please point me to the correct one). I 'assume' that I need to:
1) Create a custom res in the nvidia panel and switch it to Full Range
2) Use the colormunki SW to create and apply Colour Profile (ICM) for my TV in Windows 7 (I can see the output change at this point), but: ColorMunki SW during calibration wants me to change the Brightness and Contrast settings of the TV, do I do this on the TV itself OR via the nvidia control panel?
3) If the TV now has a ICM setup what is the purpose of yCMS in madVR and how to I extract the data from ColorMunki (or the ICM it created)

Thanks
Nathan

nvidia GTS450 --> HDMI --> Yami V2700 --> HDMI --> Pio LX608a Plasma (was professionally calilbrated a few years ago but will not doubt be off by now)

nevcairiel

22nd July 2011, 07:24

I didn't create any ICM profile.

What i did, basically:
1) Custom resolution for 24p, 50p and 60p
2) Setup Black/White point using display in-built controls
3) Measure gray-scale using ColorHCFR
4) Adjust gray-scale settings using TV in-built constrols until its much closer to the target (i have 10-point IRE settings in the TV, as well as 2-point settings - i'm using the 2 point now, it was much simpler to setup, and gives a pretty good result)
5) Confirm Black/White point are still good, if not, go back to 2)
6) Measure gray-scale and primaries with ColorHCFR
7) Enter results in madVR to create a 3DLUT

Tada!

You could add an extra point between 6 and 7, adjusting primaries in your display, if it offers those options. Any settings you can do in your display will usually be better then correcting through the 3DLUT.

Measuring in ColorHCFR was done using the AVS calibration videos running directly in madVR, which was switched to doing no calibration by itself.
I do however not know if ColorMunki is supported by ColorHCFR.

jmone

22nd July 2011, 07:43

Thanks nevcairiel - off to read about ColorHCFR and if Colormunki is supported!

madshi

22nd July 2011, 08:16

I think both are handling it correctly.
Sorry, but I disagree. We have a screenshot from the old config with a BT.709 2.6 gamma, and we have a screenshot from the new config with a BT.709 2.1 gamma (2.1 * 1.25 correction factor = 2.625), and they're noticeably different. The 2.625 image isn't just brighter, it has that grayer look to it, that the BT.709 curve always seems to produce with the newer madVR versions. I believe there must be a problem in either yCMS or in madVR.

@Fer, sorry for asking you for yet another screenshot, but could you do one with the new config with a BT.709 gamma curve and a gamma value of 2.05? That way we would have the following 3 screenshots to compare:

(1) old config, BT.709 curve, 2.6000
(2) new config, BT.709 curve, 2.6250
(3) new config, BT.709 curve, 2.5625

If (1) is right in the middle between (2) and (3) then everything seems to be fine. But I rather think that both (2) and (3) will have a different look to them compared to (1). Which would mean that either yCMS or madVR has a bug (or both).

yesgrey

22nd July 2011, 10:59

@Fer, sorry for asking you for yet another screenshot, but could you do one with the new config with a BT.709 gamma curve and a gamma value of 2.05?
Even better would be a screenshot of the old config with a 2.625 value, by using Gamma_Curve 1.0 2.625, this way we can compare exactly the same gamma curves via old and new configs.

madshi

22nd July 2011, 11:34

Yeah, good thinking.

salora

22nd July 2011, 12:03

Strange... can you try removing also the Grayscale_Measurements command? This way you should get the same picture as before, so, if you're still getting the color changes than something else might be wrong... By the way, are you using madVR v0.66?

I was thinking maybe there is something wrong in my computer settings, because I downloaded measurements from hcfr forum of my display and the gamut is normal

so here is my settings

I use MPC-HC with ffdshow video decoder output set to YV12, and madvr 0.66 set on tv display range
I have a Nvidia GT520 graphic cards set on rgb display for the desktop and set on "control by player" for the video settings
is there something wrong?
thanks

Fer

22nd July 2011, 13:44

@Fer, sorry for asking you for yet another screenshot, but could you do one with the new config with a BT.709 gamma curve and a gamma value of 2.05? That way we would have the following 3 screenshots to compare:

(1) old config, BT.709 curve, 2.6000
(2) new config, BT.709 curve, 2.6250
(3) new config, BT.709 curve, 2.5625

If (1) is right in the middle between (2) and (3) then everything seems to be fine. But I rather think that both (2) and (3) will have a different look to them compared to (1). Which would mean that either yCMS or madVR has a bug (or both).

(8) The old config A) with Gamma_Curve 1.0 2.6:

http://img194.imageshack.us/img194/9009/a1bt2600.th.jpg (http://imageshack.us/photo/my-images/194/a1bt2600.jpg/)

(9) The old config A) with Gamma_Curve 1.0 2.625:

http://img854.imageshack.us/img854/8070/a1bt2625.th.jpg (http://imageshack.us/photo/my-images/854/a1bt2625.jpg/)

(10) New config 1) with gamma processing enabled in madVR and set to a BT.709 curve of 2.05 (two near frames to falicitate comparison):

http://img684.imageshack.us/img684/7862/bt2051a.th.jpg (http://imageshack.us/photo/my-images/684/bt2051a.jpg/)

http://img836.imageshack.us/img836/7952/bt2051b.th.jpg (http://imageshack.us/photo/my-images/836/bt2051b.jpg/)

madshi

22nd July 2011, 13:54

Ok, thanks. So clearly, as I said earlier, either yCMS has a bug or madVR, or both. yesgrey, can you investigate?

alph@

22nd July 2011, 17:43

Not quite. Several LCD models still have small gamuts. It should be cheaper?...
hello yesgrey,the sony is supposed to be wide gamut http://www.neox.fr/Televions-Hi-Fi-Home-cinema/Televisions/TV-LCD/46-pouces/Sony/KDL-46HX700/KDL46HX700AEP/809386.html#_index :confused:
This is a recurrent question, so I think it's time for me to create a FAQ. I will try to do it soon, stay tuned.;)
yes,thanks it's necessary.

yesgrey

23rd July 2011, 21:07

I use MPC-HC with ffdshow video decoder output set to YV12, and madvr 0.66 set on tv display range
madVR 0.66 is not working correctly with the 3DLUT. Try with the last one and let me know if everything is working fine.

Sorry, I've missed your post when you stated the version you were using.

yesgrey

23rd July 2011, 21:08

(8) The old config A) with Gamma_Curve 1.0 2.6:
(9) The old config A) with Gamma_Curve 1.0 2.625:

Ok, thanks. So clearly, as I said earlier, either yCMS has a bug or madVR, or both. yesgrey, can you investigate?
Definitely there should be something wrong. I will look into it.

yesgrey

23rd July 2011, 21:13

hello yesgrey,the sony is supposed to be wide gamut http://www.neox.fr/Televions-Hi-Fi-Home-cinema/Televisions/TV-LCD/46-pouces/Sony/KDL-46HX700/KDL46HX700AEP/809386.html#_index :confused:
Well, it is wider than other LCD screens, unfortunately it seems to be not enough to fully cover the BT.709 standard...

The only thing that you could do is measuring the primaries for each of the colour modes the display has and see if there is any setting with a wider color gamut. If the gamut is the same in all modes so there is nothing that can be done. :(

alph@

23rd July 2011, 21:35

The only thing that you could do is measuring the primaries for each of the colour modes the display has and see if there is any setting with a wider color gamut. If the gamut is the same in all modes so there is nothing that can be done. :(
all modes give the same gamut,next time I would be careful before purchasing a display,it is very difficult to find the gamut range of a display in the net.

yesgrey

23rd July 2011, 22:32

next time I would be careful before purchasing a display
Don't worry about it, it's pretty close to BT.709. You're only missing a little of red, but those highly saturated reds usually only appear in CGI films, so it shouldn't be a big issue on your daily usage.

salora

24th July 2011, 07:53

madVR 0.66 is not working correctly with the 3DLUT. Try with the last one and let me know if everything is working fine.

Sorry, I've missed your post when you stated the version you were using.

don't worry for that

about madVR 0.67 I read that there are bugs too , and maybe in yCMS also, so it maybe better to wait for others bugfree versions?

and will it make a there a difference if I set the output of ffdshow to RGB32 or others? to my gamut I mean, because as i said earlier my display does have a wide gamut (when it's new, mine is old could be that?)
thank u

alph@

29th July 2011, 18:42

for dvd pal, what color space should I use in Colorhcfr 'Pal/ Secam or sdtv rec 601'.and what are the 'view images(red, green,blue,white) that I should use for dvd pal.

yesgrey

29th July 2011, 23:06

for dvd pal, what color space should I use in Colorhcfr 'Pal/ Secam or sdtv rec 601'.
Pal/Secam

what are the 'view images(red, green,blue,white) that I should use for dvd pal.
Could you be more specific about what are you intending to do?

alph@

30th July 2011, 16:46

thanks for your reply yesgrey

Pal/Secam

Could you be more specific about what are you intending to do?
for play dvd pal and correct gamut ,i use this 'mires' http://www.mediafire.com/?oc8zsn4llgsvyex ,(100% saturation),I found nothing else,is it good.

yesgrey

30th July 2011, 18:06

for play dvd pal and correct gamut ,i use this 'mires'
OK, I think you are trying to calibrate for each of your sources, but you don't need to do that. You only need to calibrate to one of them, and then madVR will automatically adapt each source to look accurate on your display.

nand chan

8th August 2011, 18:46

I must ask again, is there any plan of including the option of creating a 3dLUT from an ICC profile? What about .ti3 (http://www.argyllcms.com/doc/ti3_format.html) files?

I need this because my profiling software (ArgyllCMS) only outputs these two types of file to hold profiling information, from which I want to create a .3dlut to work with madVR.

Mr Alpha

8th August 2011, 21:24

Mark_A_W

8th August 2011, 22:44

Hi madshi/yesgrey

I'm trying to get the new version working.

I have deleted and let madVR re-download yCMS.

- With 3D LUT created with previous version, I get an error that the 3D LUT does not contain primary colours info (yes it did!).

- After deleting the old 3D LUT, madVR is unable to create a new one. I get a "Creating the 3DLUT for the display blah blah failed" error.

The measurements are the same dodgy ones I had before - for testing only (I'm testing the robustness of your system ;) )

Primaries:

red, Yxy, 5.112, 0.628, 0.337
green, Yxy, 9.018, 0.294, 0.595
blue, Yxy, 1.672, 0.156, 0.078
white, Yxy, 10.833, 0.321, 0.342

Greyscale:

20, Yxy, 0.301, 0.338, 0.349
30, Yxy, 0.179, 0.316, 0.333
40, Yxy, 2.705, 0.311, 0.335
50, Yxy, 5.037, 0.311, 0.340
60, Yxy, 7.476, 0.315, 0.340
70, Yxy, 10.364, 0.320, 0.341
80, Yxy, 10.603, 0.320, 0.341
90, Yxy, 10.749, 0.321, 0.342
100, Yxy, 10.850, 0.321, 0.342

Any ideas? Do you want me to activate debug mode.bat? (How do you deactivate it?)

Fixed it.

Should be:

30, Yxy, 1.179, 0.316, 0.333

However I'm still getting the fluro greens happening in white. I will look for another bad value.

nand chan

8th August 2011, 23:14

I found making the 3dLUT from the ti3 file created by ArgyllCMS by hand using madVR's interface relatively trivial. You can open the ti3 file in a texteditor of your choice and then just copy/paste the results.

Oh? Is there something I'm missing here? Where would I insert this information and how?

Afaik madVR's yCMS interface only allows me to insert the primary coordinates as well as a grayscale ramp.

How do I parse a full LUT?

Also, I guess I could find out the structure of madVR's .3dlut files, then write a custom interpolation engine + YUV convertor, but it would probably be nowhere near as clean. I suppose I could use the same libraries yCMS used for interpolation.

cyberbeing

9th August 2011, 00:58

How do I parse a full LUT?

You don't, since measured Primaries + Whitepoint + Grayscale is all yCMS currently uses for creating its LUT. Supporting more measurements has been something on yesgrey's todo list for quite awhile.

As I've stated previously, if anybody with the necessary coding knowledge can hack at ArgyllCMS colorprof to output a 3DLUT using modified ArgyllCMS XYZ LUT AtoB/BtoA tables as a base, that would be optimal from a calibration workflow standpoint. It should be easier now that madVR 3DLUTs are working in RGB->RGB, but that could change back to yRGB at some point. In any case, there is no longer any need for a 3DLUT doing YUV->RGB when using madVR.

yesgrey

9th August 2011, 01:01

However I'm still getting the fluro greens happening in white. I will look for another bad value.
Your primaries measurements are strange... The R,G,B luminances sum should equal the white luminance, but in your case they are very different...

yesgrey

9th August 2011, 01:07

I must ask again, is there any plan of including the option of creating a 3dLUT from an ICC profile?
It's on my todo list, but I don't have any idea when I will have time to do it.

Mark_A_W

9th August 2011, 01:48

Your primaries measurements are strange... The R,G,B luminances sum should equal the white luminance, but in your case they are very different...

Thanks yesgrey, I will check them again. Remeasure if I don't find any mistakes in transposing the numbers.

I have an old CRT projector....it has some quirks. I am a test case for the robustness of your software...

But it still looks fantastic - only in the last year have digitals started to perform better than my 14 year old CRT (and they STILL can't do a full fade to black).

nand chan

9th August 2011, 02:17

You don't, since measured Primaries + Whitepoint + Grayscale is all yCMS currently uses for creating its LUT. Supporting more measurements has been something on yesgrey's todo list for quite awhile.

As I've stated previously, if anybody with the necessary coding knowledge can hack at ArgyllCMS colorprof to output a 3DLUT using modified ArgyllCMS XYZ LUT AtoB/BtoA tables as a base, that would be optimal from a calibration workflow standpoint. It should be easier now that madVR 3DLUTs are working in RGB->RGB, but that could change back to yRGB at some point. In any case, there is no longer any need for a 3DLUT doing YUV->RGB when using madVR.

Ah, right, I see what you mean - yeah, that's what I'm currently doing - measuring the primaries manually and inputting them + the grayscale ramp to yCMS.

Also, what's so hard about using YUV->RGB instead of RGB->RGB? Isn't the YUV color space mathematically identical to the RGB color space? Conversion is relatively straightforward and well-defined (http://en.wikipedia.org/wiki/YUV#Conversion_to.2Ffrom_RGB)

All you need to do is create a regular 3dLUT for sRGB->XYZ (which can be done mathematically) and change each R/G/B entry to Y/U/V using the described formulae (resulting in a YUV->XYZ), then go through each entry and find the closest match to the given XYZ in the monitor's interpolated output profile (as derived from .ti3 as RGB->XYZ), and write that R/G/B value into the final stage of the YUV->RGB 3dlut.

You don't even have to bring the grayscale into the equation, unless I'm mistaken.

I don't quite know how to do the correct interpolation myself, especially not using derived gamma functions.

Finding the nearest match in 3d space would also potentially pose a large performance problem when dealing with 1.07 billion entries, especially if using an O(n²) algorithm and using trigonometric distance functions to sort (though that can be skipped for the vast majority of entries by simply filtering out values which deviate too far in at least one axis).

Though, if you have the memory for it, I suppose you could create a three dimensional BST (that is, a BST containing BSTs containing a BST each) and use a recursive algorithm to find all points that are likely to be matches.

How does ICC do it, either way? If you want an easy way out, you could use argyllCMS's collink.exe utility to link the device profile to the virtual sYUV profile (as calculated from sRGB), then just map each value from 0/0/0 to 1023/1023/1023 through it. This way you can also take care of things like rendering intents (some people would really like perceptual mode instead of absolute colorimetric).

Ps. How deep is the bit-depth of madVR's 3dLUTs either way? 8 bit? 10 bit? 16 bit?

A nice 16 bit YUV -> RGB conversion 3dlut would be the ideal solution, with post-dithering down to the maximum bit-depth supported by the graphics card, which is unfortunately only 10.

It's on my todo list, but I don't have any idea when I will have time to do it.

How about an option to format a .3dlut file from a text list of R/G/B -> R/G/B or Y/U/V -> R/G/B entries? That way I can implement the logic myself by using existing ICC libraries.

nand chan

9th August 2011, 15:45

Hmm, whenever I add the grayscale info from the .ti3, my image looks up very distorted afterwards and I doubt it's supposed to look that way.

For comparison:

Before (just gamut correction) (http://img840.imageshack.us/img840/7815/beforem.jpg)
After (gamut correction + grayscale) (http://img847.imageshack.us/img847/7430/afterm.jpg)

Grayscale info I used (http://pastebin.com/KpwAPGVf)

Also, I left the ICM profile generated by ArgyllCMS as default in the Windows color management - I figured this would be correct since the profiling was done while the profile was active, no?

Edit: Solved this problem already, was using the wrong values