AVS Forum banner

Color Cube (or 3D LUT) vs. more traditional calibration.

15K views 74 replies 21 participants last post by  Steve_Shaw 
#1 ·
What is the general consensus (if there is one), on color cube calibration when compared with traditional calibration?
 
#2 ·
a 3D LUT gives you infinitely more control for correction.

Assuming that you've got enough bit depth in the 3D LUT and the path to the display, you'd always choose a 3D LUT for quality.

Now you need software to create a 3D LUT. With standard CMS controls, you could calibrate with a pattern disc and the x,y,Y numbers off the back of a meter.
 
#3 · (Edited)
Traditional calibration involves using only the controls your display has to get things as 'good' as they can be. This means you're limited by the original design of your display and that includes all it's inbuilt weaknesses. On a PJ like the AE8000 you'll probably have a CMS with a single point for each color and a 9 or 10 level advanced menu to better adjust gamma and greyscale balance. That's all you can calibrate with and your end result will be as good as you can go with these controls.

3D Luts change this entirely and allow you to get around most or all of the original design limitations in your display when it comes to calibration. You get 21 point greyscale instead of 9-10 points in your PJ. You get hundreds or thousands of points of individual color calibration instead of the 6 in your CMS. With a 3D Lut box you can take a display that has an average picture post calibration and make it into a near reference display. With a traditional calibration you might notice that red at 50 IRE looks oversaturated and green looks undersatured and you'd never be able to adjust that out. With a 3D Lut that all goes away.

I have an AE4000 PJ that I'm using a Radiance Mini 3D on I've auto calibrated. The measurements in the before and after charts is stunning, and the picture quality is something you wouldn't think is possible with this level of PJ.
 
#5 ·
I don't know much about 3D LUT however, I have a question.

If you want to use the LUT to correct the display, how do you first maximize the display for all types of dynamic range so that the LUT doesn't clip in places making it less useful?

=Brian


You need to use the displays controls to calibrate white level to d65 and if possible select a gamut that is wider than the one you are calibrating for.
 
#6 ·
OK.... What about gamma ... I have no trouble getting my projector to have a nearly ruler flat grey scale but I don't think I ever got the gamma right so far.... Maybe it's due to my screen or my completely pitch black room.

Also,... do the cheap EEcolor boxes ($50 on fleabay) work as LUT holders?... I'm thinking no or the more expensive models wouldn't be selling for $650...

I see a there is a beginning of support for LUT in XBMC and I'm happy about that because XBMC has become my favorite source.

-Brian
 
#7 ·
OK.... What about gamma ... I have no trouble getting my projector to have a nearly ruler flat grey scale but I don't think I ever got the gamma right so far.... Maybe it's due to my screen or my completely pitch black room.

Also,... do the cheap EEcolor boxes ($50 on fleabay) work as LUT holders?... I'm thinking no or the more expensive models wouldn't be selling for $650...

I see a there is a beginning of support for LUT in XBMC and I'm happy about that because XBMC has become my favorite source.

-Brian

I leave gamma at defaults, all that is recommended is to get reference white level to d65 and gamma is done through the calibration software, if you do have difficulty with gamma you could try using a setting that is closest to your target.


I think you are referring to the domestic eebox which is no good, you need the commercial box which is ~$600 new.
 
#8 ·
OK thanks,...

I'm far behind the curve as it were. I don't want to throw $600 at this part of my system just now.

I imagine cheaper options will surface.

I was glad to see a lot of effort has been made to work support for 3DLut into XBMC and I think I'll dabble with what's been done and see how it goes.

-Brian
 
#9 ·
OK thanks,...

I'm far behind the curve as it were. I don't want to throw $600 at this part of my system just now.

I imagine cheaper options will surface.

I was glad to see a lot of effort has been made to work support for 3DLut into XBMC and I think I'll dabble with what's been done and see how it goes.

-Brian

Ok firstly you don't have to need the eebox if you are running things on your PC as the PC can hold the LUT anyway!


The eebox is if you want to hold the LUT "OFF" PC for when you want to run separate sources through it eg. Satellite box/bluray player/PC etc.. (through) eebox (to) display!


Have a look at this thread which shows you how to create a 3d lut using Argyll/dispcal Madvr on a PC and all software needed for free and then all you need to do is find out how to incorporate the lut into XBMC, which I think is very similar!


http://www.avsforum.com/forum/139-display-calibration/1471169-madvr-argyllcms.html
 
#10 ·
Thanks for the XBMC tip, looks interesting. You do know that you can do all that with MPC+madVR (including the test pattern generator) right? That would allow you to play around with 3dluts for free until XBMC incorporates them.
 
#13 ·
So far my home theater has no windows...

Actually that's not fully correct... I have a laptop I bought just for HCFR and I am loading it up with Argyll and DispCal and so on... But the only source that could possibly make use of a 3DLut I would have for now would be a Chromebox running XBMC.

I don't have a decked out PC running to my HT or I may know more about MadVR and so on...

Still.... Looks like XBMC has (experimental) support for the 3DLUT and I may be able to give it a whirl.

-Brian
 
#14 ·
madVR is an incredibly good rendering package but you need a dedicated and powerful GPU to make the most of it. With my intel HD3000 graphics on the laptop I can only run the basic upscaling algorithms plus dithering, it drops frames with the 3DLUT turned on.
 
#18 ·
Thanks Ted... I gathered that.

I sort of think it may make sense to dial in the traditional calibration as well as possible using a display's controls before using a 3D LUT so that the 3D LUT has less work to do...Of course,.. just making sure the gamut is wider than standard if and when possible.
 
#19 ·
There is a good 1D/3D Cube Resolution Comparison Guide for the users to visualize the difference of calibrated color points precision over various cube resolution sizes vs. the calibrated color points using the traditional internal display controls calibration of 1D LUT (Grayscale/Gamma) + CMS.
 
#21 ·
There are many more that will tell you that every display benefits from a LUTs calibration.

The minority say that it depends on your display. If you can get perceptually accurate image with a display's CMS, then what's the point of buying the additional equipment and running fairly long LUTs calibrations to only get prettier charts.

A sound approach is to get the display as accurate as possible with the internal CMS and then run a large color checker program to find the errors and whether or not that they're likely to be seen. If they are strongly visible, then pursue a LUTs calibration.
 
#22 ·
A sound approach is to get the display as accurate as possible with the internal CMS and then run a large color checker program to find the errors and whether or not that they're likely to be seen. If they are strongly visible, then pursue a LUTs calibration.
And the dirty little secret is that there can be rather "huge" errors, while still maintaining a perfectly acceptable image.

In fact, I'll go so far as to say LUTs are probably severe overkill in the home environment ... for a laundry list of reasons. ;) OTOH, in the professional environment one would expect displays to be a close to "perfect" as possible. That being said, if you got the money to burn, knock yourself out. :)
 
#24 · (Edited)
I hate posting camera shots of displays but those below do capture the main features of what I see in an A/B comparison of an image calibrated using the internal CMS vs. a LUT. In this case it's an F8500 CMS vs. a 65^3 eeColor LUT.

The measured color difference errors for both cases are quite good with 88% of a random 1000 test set < 1.5 dE2K for the CMS and 99% < 1.5 dE2K for the LUT.


The LUT cleans up a bit of undersaturation compared to the CMS but you can only see this difference in a side by side like this. For example a color picker sample of a flesh tone from the top image yields and RGB triplet of [100,65,60] and the identical pixel from the bottom image yields [100,62,56] which has a color difference (dE2K) of 1.32. The color difference error of the yellow in the butterfly between the two images is 0.71 dE2K and 0.67 dE2K for the darker part of the red hair.

EOS 7D images of a still from the "Tree of Life" BD at identical exposure settings.


Other displays may not calibrate this well using the internal CMS so you will need to evaluate it's capabilities first before deciding, and if you are so inclined you can test-run the benefits of a LUT for free via HTPC and madVR. The other main benefit besides color correction is the ability to easily switch between LUTs with different transfer functions and/or HD/SD/other colorspaces.
 
#25 ·
I hate posting camera shots of displays but the below does capture the main features of what I see in an A/B comparison of an image calibrated using the internal CMS vs. a LUT. In this case it's an F8500 CMS vs. a 65^3 eeColor LUT.
Hi Zoyd, I just though an another way to evaluate the Internal CMS vs. LUT using any Picture as a reference using LightSpace LUT Preview feature.

Start by calibrating the display perfectly with the available calibration controls.

To this calibrated display run any patch set of grid-based (17x17x17) or with custom color points using Argyll or LightSpace to this already pre-calibrated display.

Export the correction LUT in 65-point eeColor format TXT file and open LightSpace, go to Tools -> LUT Preview.



The LUT Preview Menu opens a standard navigation window that enables the user to select any image to preview the active LUTs within LightSpace.

Load any picture you like (TIFF,TGA,PNG etc.) that you will use as a reference.

When the image is selected a new LUT Preview window will be displayed.

Click to a '+' tab and it will ask you to import any LUT format file (like eecolor 65.txt) after that it will apply the LUT you loaded changes to the image, and you can swap the tabs to see the differences. ;)

This comparison will show if there differences that can be noticed visually by the user.
 
#28 ·
Sure, no one needs to do any of this to enjoy their TV, but the original question was how does current display based calibration compare with LUT based, and I offered one example for context.
 
#30 ·
And my point is that if one has to "squint" to see the difference between three different conventional calibration results (albeit with the stipulation that the displays are *not* directly compared,) then how much of a real impact could a LUT vs conventional calibration make when the target display is fed actual live moving pictures (as opposed to still frames or test patterns?) :)

In short, the real advantages to the LUT approach are:

1) The semi-automatic process.

2) 'The ... the ability to easily switch between LUTs with different transfer functions and/or HD/SD/other colorspaces.'

With 3) 'Improved color corrections' bringing up the rear.
 
#29 ·
Here is a quick example how LightSpace LUT Viewer works:



Accoding to this way of comparing 1D/3D LUT Calibrations the problem is that by trying to profile a calibrated display, this is often not the best way to make a correction 3D LUT, as you need the gamut to be larger than target, to allow the LUT some 'space' to work.
 
#31 ·
Accoding to this way of comparing 1D/3D LUT Calibrations the problem is that by trying to profile a calibrated display, this is often not the best way to make a correction 3D LUT, as you need the gamut to be larger than target, to allow the LUT some 'space' to work.
And if one has a display that's wider than R709 on two sides of the gamut but narrower on the third??? And by narrower, I mean ~75% sat is the best you can reach for cyan and green.

Even with that "huge deficiency," I've yet to really notice it on actual HDTV programming or BD, because I'm too busy watching the movie or TV program ... If I start noticing the display issues, that's just a sign that the movie a/o TV programming quality needs to be upgraded. :D ... And so we come full circle ...
 
#32 ·
I don't disagree that on some displays the color correction aspect is minor (the above was an example of this) but for others it may be more pronounced, that is why I said:

Other displays may not calibrate this well using the internal CMS so you will need to evaluate it's capabilities first before deciding, and if you are so inclined you can test-run the benefits of a LUT for free via HTPC and madVR
I'm sure there are examples where color correction will be the primary benefit. Ted's example would easily be detectable as a standalone correction (still or moving), although I don't know what display/situation he is attempting to emulate.
 
#33 ·
Ted's example would easily be detectable as a standalone correction (still or moving), although I don't know what display/situation he is attempting to emulate.
The picture I used as a reference and the loaded 3D_LUT_1 correction are just example to show the way the LUT Preview of LightSpace can be used for these comparisons.

3D_LUT_1 was a cLUT I had randomly loaded to LightSpace for example only, not emulating any display 1D vs. 3D LUT difference.
 
#44 ·
Since I brought it up, the costs for me between one and the other are not significant. I was purchasing a Lumagen for my projector already. I have had a couple TV's calibrated in the past with normal methods, and wasn't really blown away by the improvement, and questioned the value of having it done. I was offered calibration services on the projector by the person who sold me the Lumagen, and they said they could calibrate it in conjunction with the Lumagen too. My choice was between having no calibration done, having someone local do calibration to the projector (without the LUT calibration on the Lumagen), or shipping my projector and having the calibration done with the Lumagen. It was basically a question of whether 1. Calibration at all would be worth it, and 2. If having the 3D LUT calibration with the Lumagen was worth the costs to ship the projector.
 
#45 ·
Since you already have the Lumagen. I would consider option 3: Go ahead and buy a meter and software and do the calibrations (and LUTs if you want) yourself. In your home, on your screen.

Calibration is always worth the effort, even if the results may not be what one might hope for. In this case, the main problem with shipping everything off for calibration is the screen factor ... the screen (and other room issues) will effect the results, so unless the calibrator can duplicate your environment exactly, the results will be sub-optimal and will probably need some "re-correction" once everything is reinstalled. So in short, you might as well DIY. Cost wise, you appear to be 2/3rds to 3/4ths of the way there anyway.

Good luck ... and apologies for the side-show. :)
 
#53 ·
I agree with you Ted. What you are showing makes it very clear that even what is thought to be a acceptable (below 3) dE 2000 can be seen.

It has become very clear to me that you want your Avg dE2000 to be below 1 and your high dE2000 error not to be greater than 2.25 dE2000 for a Plasma, because it probably will shift.

Looking at zoyd's reference pictures (Both CMS & 3DLUT), the flesh tones was ok although a bit blotchie in the cheek area , but what really stands out are the blue streets. The correct street color has always bugged me, and something I look close at when running a Blu Ray disc. Of-course this is just one of many examples of color & luminance error.

The only way I have found to correct these errors is by doing a very large number of custom triplet patches. Of-course saying you have everything in your pre profile setup to optimize your profile. That doesn't mean you want to do a full grayscale or CMS using the controls of your display, that type of pre calibration you want to keep to a min.
What imho you want to look closely at is what mode you want to use, Color Temp, Color Gamut, contrast, brightness, color/tint setting and 100% White Point.

One other factor in getting the holly grail profile/calibration, is in making your meter profile matrix (if using a color and spetro meter) and how accurate your source and reference meters are.

On a side note dispcalGUI powered by argyllCMS also has a very good color measurement report that shows what the triplets color should be and what they are measured at.

ss
 
#57 ·
In conclusion. Worrying about an average color dE of 3.0 vs 0.7 is a bit like putting a band-aid on a paper-cut while ignoring that the patient is bleeding out from a severed limb. :)

Perspective. :)

That having been said, there are certainly several other worthwhile advantages to the 3DLUT approach. It just turns out that "absolute" color corrections may not be the most interesting or useful part. Much like the "4K resolution" of 4K/UHD may not be the most interesting or useful part of UHDTV. :)
 
#58 · (Edited)
Worrying about an average color dE of 3.0 vs 0.7 is a bit like putting a band-aid on a paper-cut while ignoring that the patient is bleeding out from a severed limb. :)
I have to agree that in general, it's going to be hard to distinguish between DE 3 and DE 0.7 - to reliably do so, typically requires side by side comparison. Under side by side conditions when viewing neutrals, it is even possible to distinguish a 0.7 DE error though.

So in summary, this is what I understand to be the situation: viewing a display on its own (relying on memory), DE's of 3 are probably going to be very hard to pick up, unless they fall in such a place as to cause an obvious artifact (such as grey scale gradation errors). On the other hand, if one is striving for the highest possible fidelity, and/or the situation is critical (you are mastering content), then why wonder about the effects of a 3DE error when you have the means to reduce them to the point where even a side by side examination cannot tell them apart ?
 
#59 ·
there's also another wrinkle to perhaps consider: yes, we have remarkably adaptive sensory systems, and a grayscale that is off by a little bit might not be a major issue if you're primarily watching content on that display. But adaptation will not come to the rescue if the grayscale isn't consistently off by the same amount along the neutral axis, for example.
 
#60 ·
But adaptation will not come to the rescue if the grayscale isn't consistently off by the same amount along the neutral axis, for example.
That's a good point that I'd certainly agree with. Only thing I'd add is that not only the amount, but also the direction plays a role. E.g. 1.0 dE "too red" consistently over the whole neutral axis may be ok, but 0.5 dE "too red" at certain points and 0.5 dE "too green" at other points (which amounts to the same 1.0 dE "range") maybe not.
 
#70 ·
CalMAN ColorChecker Now Includes 3D LUT Visualization Features

See how accurate your monitor could be with new virtual 3D LUT capabilities and a free LUT visualization tool.

Now you can do more than test the accuracy of your display. Create a virtual 3D LUT in ColorChecker to see the difference a calibration could make! CalMAN can now create a virtual correction matrix unique to your monitor's color performance so you can see just how good your display could be. Once you've created the virtual 3D LUT in ColorChecker, upload your own photos and images to CalMAN Eyeglass and see the before-and-after effects of your corrective 3D LUT.

Download CalMAN 5.4 ColorChecker (Free Version) to get started.



(Above: photo viewed in CalMAN Eyeglass with corrective 3D LUT uploaded. Photo is rendered in split screen with the native display output on the left, and the corrected output on the right. This image is for demonstrative purposes only and will appear differently on different displays.)

Download CalMAN ColorChecker QuickStart Guide
 
#71 ·
Hi guys,
I've been trying to read a bit more on 3D LUT, but it's much too technical for me. I'd really appreciate your advice here in very simple layman's terms: I own a Sony VW520ES (VW665) and I am really impressed with picture quality as it came already with factory default setting. But I was recently thinking about having it calibrated...

The guy that I'm speaking to about calibrating my Sony is suggesting that for such "high-end" machine, he would recommend 3D LUT to get most out of it. But I'm wondering if this is not maybe an overkill? Would I really see such an improvement, despite this projectors being one of the best on the market?

I would really appreciate some advise whether it makes sense to invest a bit more and
(A) have my Sony 520ES calibrated with standard calibration tools etc.?
(B) have it calibrated with 3D LUT calibration and buy the eeColor 3D LUT Box?
(C) let it be like it is - I still find its picture quality phenomenal, but it is true that my previous PJ was years old, so can't compare really...

many thanks!
 
#72 ·
I would really appreciate some advise whether it makes sense to invest a bit more and
(A) have my Sony 520ES calibrated with standard calibration tools etc.?
(B) have it calibrated with 3D LUT calibration and buy the eeColor 3D LUT Box?
(C) let it be like it is - I still find its picture quality phenomenal, but it is true that my previous PJ was years old, so can't compare really...
The EEcolor cannot pass 3D, HDR or 4K, so bear that in mind if choosing the EEcolor as your 3DLUT holder. I think most people looking to use 3DLUT on a device like yours would go for a Lumagen Radiance Pro, but if you only watch 2D bluray or lower then it can be a good solution.

A post-calibration report on your unit without a 3DLUT, profiling it across a large number of points, would give you numeric data (delta Es, or dEs) telling you objectively how far your projector is away from reference. dEs above 1 are generally considered noticeable, and if you have a high percentage of those then you will be watching with colour which is noticeably not reference.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top