PDA

View Full Version : Does anybody use "Display Color Space" LUT with success?



madno
02-21-2013, 04:23 AM
Hi,
does anybody know what the "Display Color Space" in CS settings is intended to do?

My assumption is, it should compensate the "wrong colors" a display shows. The information how wrong the colors are, is taken from the ICC profile (the LUT) made for that display.

Unfortunately LW seems to do this in a way that differs from all other applications I use.

I tried now in LW 11.5 and, like in all versions before, LW shows shades of gray with a shift to pink (most noticable in light gray / near white shades).
My ICC profile is made for a NEC 271 SpectraviewReference (wide gamut, hardware calibrated). Meassurement device was a basICColor discus colorimeter (high end device).
Using the Profile in PS, Illustrator, Mozilla, XNView etc. gives perfect results. Super accurate whitepoint (D65), super accurate gray balance (no banding, all shades in all the test stips are distinguishable, no hint of pink).

In LW the sRGB setting in "Display Color Space" do not show this pink shift.

111767111765111764

jwiede
02-21-2013, 10:14 AM
I believe you're seeing an ICC profile version issue. I vaguely recall this coming up when CS was first added, and in various 10.x patches, and the outcome was something like "LW only supports up through v2* of ICC profile, using higher gives odd results". If search worked, searching on 'Lightwave ICC "pink cast"' ought to bring up the older thread, unless it was in a private area (can't quite recall). I'll try to find it this evening, but with search as borked as it is, might just be easier/faster to try outputting a lower version of ICC profile from the basICColor software and see if it works better in LW.

*: I don't recall what LW's "ceiling version" was for ICC profiles in the thread precisely, but vaguely recall the thread had it accepting v2 and failing on v4. Worst case, if v2 still causes the issue, try v1?

madno
02-21-2013, 10:48 AM
Hi Jwiede,
thanks for chiming in. I checked the ICC version topic already. LW does not load V4 profiles at all. So I made a V2 version. But even with V2 priofiles this pink cast should not appear. At the moment I just can say, without a manual describing what the display ICC is meant for, I am lost. The only thing I am sure about is, that it does not do what a display profile does in all the other apps with a color management feature for displays.
Another thing I noticed is, that the "Output Color Space" and "Color Space" settings in VPR are linked, which makes sense to me. So I come to the conclusion that everyting seems to be as it should be, only the last link in the chain - the display ICC profile - leaves a big question mark.

[Edit]
Regarding the "forum search"
I searched using a lot of different string patterns about the topic. The result was always "nothing found". First I thought, hey, I can't be the only one fighting with CS in LW. Then I noticed, that I was not logged in (where in my forum profile is the settting to keep me logged in more than some minutes?). And guess what, seems to me, if you are not logged in, search always says: "no result". If I am right about this, it Would be nice to get the hint, that only logged in users can use the search function. Did you say something about "borked" search?

Arghhhh, while I was writing, I was logged out again
111779.

My habit now is to either write a forum entry in an editor and copy / paste it into the forum, or write it in the form and copy / paste it to an editor frequently (in case I get logged out).

gerardstrada
02-21-2013, 07:50 PM
Notice a display ICC profile is not a 'LUT'. That's other format. LW uses LUT format and LW color tables for colorspace corrections. Guess when LW load a display profile it should be converting from ICC to a LUT, and the thing is that we can't really translate an ICC profile to just raw primaries - too much information is missing.

Guess LW is doing wrongly this behind the scenes because on the surface, it seems like ICC profiles contain primaries and white-point, but it's not that simple. An ICC profile does not have raw primaries - it has the matrix of values necessary to transform values I/O of the PSC (XYZ or LAB), but does not actually specify the raw primaries like in a LUT profile. The values in the matrix include the chromatic adaptation function (adjusting between white-points mainly).

LW inded needs an improved native support for ICC profiles. 'Til then, you might want try ICC profiles within LW with the SG_CCFilter - part of the SG_CCTools:

http://www2.informatik.hu-berlin.de/~goetsch/CCTools/

(Would recommend to make the colorprofiles.txt manually)



Gerardo

madno
02-22-2013, 07:39 AM
Hi Geradstrada and others,

the CCTools seem to be very interesting.

I am now searching for the CCTools tutorial that must have been on Lightwiki in the past
(http://www.lightwiki.com/SG_CCTools_-_For_Color_Management_and_Linear_Workflows).
Unfortunately it is not available there anymore.
Has anybody an alternative link or know another way to get that information?

Thanks in advance.

[EDIT]
Some issues of HDRI3D magazine where mentioned as well with regard to CCTools.
But I learned its gone. Do you maybe know how to get the articles from issue 18 and 19. Maybe a site to buy a pdf download or something like that?

gerardstrada
02-22-2013, 02:00 PM
Hello Madno,

You can find the same documentation of the lightwiki article in this thread:

http://forums.cgsociety.org/showthread.php?t=617417

I haven't published the old HDRI3D articles in lightwiki because there's an incoming article with improved workflows which uses a new version of the SG_CCTools (and other free applications) that offers much better results.

Regards,



Gerardo

madno
02-23-2013, 01:55 AM
Thanks for the link.

New SG_CCTools and other apps, new workflow, better results?
Can you give any hint about whats coming, timeframe?

I am not a professional LW user, just need it sometimes to make marketing stuff for our company. But even though I am normally not on project deadlines, like a lot of you are, I already feel that I loose to much time in guessing colors. Ok, gamma curves are not a problem as LW let me choose what an how to handle that topic. But colors in relation to the differents gamuts are guesswork for me.
In PS etc. its:

set color settings correctly (including profile for the wide gammut display)
activate softproofing (e.g. for ISO newspaper)
choose colors for the image
like what you see
save image converted to desired color space (or embed the ICC profile)
send file to print house
get the result that looks nearly the same like on the display.


In LW (and the other 3D tools I know), its always something like:

set gamma curve (real ICC workflow seems not to be implented)
choose colors for the image
like what you see
render and save image
open in image editor and simulate output color space (e.g. ISO newspaper)
notice that the cool saturated colors you saw on the wide gammut display got compressed into the smaller ISO newspaper space in an undesired way, other colors shift as well (preceptual rendering intent) or banding issues appear (colorimetric intent)
go back to LW
change the surfaces
and so on


Maybe there is a chance of softproof functionality coming?
That would be impressive for LW.
By the way I saw one thread where you showed how to use the CCnode to get some softproofing on a per surface basis. Quite cool, but a lot of work if there are many surfaces. An one would need to change each node to simulate another output space.

Ups, that became a long post, but forgive me, I am really curious about what might come.

gerardstrada
02-23-2013, 06:08 PM
Yep, it's a new workflow that takes into account aspects that have been overlooked for many years now. If these aspects were integrated in 3D and compositing packages, it would be far more easy for us get realistic results in VFX and animation.

The hard part is adapt an in-house pipeline to a public pipeline without the user has the necessity of spending on new commercial toolsets. This implies that the new tools for his/her common pipeline have to be only free tools. A LOT of tests!

Would recommend two things for the first part of your procedures... Keep in mind that the working colorspace should - as possible - contain ALL colors of the output colorspace. sRGB could not be a good choice if you go for print. Consider also that some device-dependent color spaces (printers) are not 'well behaved' for soft-proofing and not all are well optimized for all rendering intents. Then, you may experience bandings and cross-channels. A good strategy in such cases in convert first to a standard CMYK color space and later convert from there to the printer profile with a device-link profile. You should get better results in that way.

For the second part of your procedures, the soft-proofing functionality is already there in LightWave thanks to the SG_CCTools:

http://forums.newtek.com/showthread.php?127209-CYMK-Ouch&p=1243952&viewfull=1#post1243952



Gerardo

madno
02-24-2013, 06:27 AM
I read the threads and, like you suggested there, sent an email (and donation) to SG letting him know I am interested ;-)

By the way my working space is normally AdobeRGB.

Just for fun:
AdobeRGB (solid) and my display (wireframe)
111906

AdobeRGB (wireframe) and ISOnewspaper26 (solid)
111907

AdobeRGB (wireframe) and ISOcoated (solid)
111908

Its a pity that all those beautiful colors I see on my display can't be printed ;-)

gerardstrada
02-24-2013, 07:51 PM
Newspaper color gamuts are pitifully very very narrow as you say. Guess your best bet might be keeping the perceptual relationships of colors. In such case you would go from aRGB to a more compatible CMYK space with Perceptual rendering intent and from there to the ISOnewspaper with a Device-Link profile :)

The colors will be lost anyway, but at least the final result will resemble better the intended color appearance.



Gerardo