or Connect
AVS › AVS Forum › Display Devices › Display Calibration › A comparison of 3DLUT solutions for the eeColor box
New Posts  All Forums:Forum Nav:

# A comparison of 3DLUT solutions for the eeColor box - Page 5

Quote:
Originally Posted by Light Illusion

You really do have to treat all colours equally, or you will never be sure of the final calibration.

Quote:
Originally Posted by Light Illusion

Display devices are not perceptual, they treat all colours the same - and as it is the display we are calibrating, not the uses perception of the display, that is what is important.

You do realize that the very term "color" is itself a perceptual phenomenon. And as Graeme has pointed out, there is no single "objective" space that represents colors.

So by treating colors equally in device space coordinates, you are not necessarily treating them equally in any number of other spaces.

By your logic, we should quantify calibration error in device space coordinates, and nobody does this. Delta E is used for very good reason.

Just so we're on the same page with delta E, here's another question:

What do you think the relationship between delta E and perceptual uniformity is?

### AVS Top Picks

Quote:
Originally Posted by gwgill

But why do you even think the equality of the samples in any space is a desirable property of a sample set ?

Hint - it's not. The goal is actually to reconstruct as visually accurate a model of the device behavior as possible from a sample set of a given size.

Interesting. So you build a model of device uniformity, and incorporate that into the sampling algorithm. Am I on the right track?
This thread is interesting... in the spirit that a reduce patch size can yield same or better result, how about something in the order of a 5x5x5 like in my old radiance, so I don't have to get the eecolor box and save some \$\$\$ ?
Quote:
Originally Posted by sillysally

What I am talking about is when doing a standard 10 point GS and a 6 point CMS. How you go about getting to the point you have good dE's, gamma ect. Not what your charts look like, what are you doing with the internal settings is what I am talking about.
Just because you can get good looking dE's, gamma, gamut, doesn't always mean you are getting the best performance from that display.

imo the charts I posted for Graeme's patch set using LightSpace are very good looking and show great results. However I have better performing LUT's .that don't have as good of looking charts and dE's.

Why do folks post when using say CM auto cal, that the dE's and charts look good (well within the 3 dE limits) but there still is a problem.

What did the client say, was there a Lumagen 20XX involved in this LUT, from were did this patch set come from. However all that makes no difference because we are talking about a VT60 and what your claim is.
How you get these is pretty much irrelevant. The only thing that's important is the result. You have this exactly backwards. You are elevating the method of adjustment over the results of that adjustment. On the the Sony 1000, the method for color correction is, well, doing nothing. On the VT60 and Zoyd's D8000 there is no LUT at all. The method is to use the display's internal CMS along with the best Picture preset and then just measure the results inside the color space in any way you like.

The only way anyone can have a public discussion about this stuff is to use objectively communicable data. Simply asserting that a result is "better performing" because. . . . why? is not informative to anyone other than the speaker.

If there is a problem with a calibration result then one should be able to describe it, and if you can describe it, then it should be measurable. The alternative is to fall down a rabbit hole of feelings and subjective responses that are notoriously subject to all sorts of biases, in particular confirmation bias.
Quote:
Originally Posted by spacediver

Quote:
Originally Posted by gwgill

But why do you even think the equality of the samples in any space is a desirable property of a sample set ?

Hint - it's not. The goal is actually to reconstruct as visually accurate a model of the device behavior as possible from a sample set of a given size.

Interesting. So you build a model of device uniformity, and incorporate that into the sampling algorithm. Am I on the right track?

Every model has a number of free parameters and a form that expresses those parameters in the detailed behaviour. An extreme example in one direction is the built in controls in a display - done well only a handful of parameters may be enough to match the desired behaviour quite accurately.

A larger the number of parameters can be much more accommodating of a wide range of device behaviours, but more parameters means that more information is needed about the device behaviour to setup the model.

Any model with a tractable number of parameters has to make assumptions about the continuity and locality of the device behaviour - to do otherwise leads you nowhere useful.

So the most efficient test sample set is the smallest one that can be used to set the parameters of the model accurately. Real world considerations mean that the ideal set is probably not the maximally efficient set - there are consistency limits to instruments and devices, and the test set is much more robust and useful if it doesn't make assumptions that are only true for one particular device model.

The results of using LS with the ArgyllCMS test set seem to indicate that it is a reasonably robust set, even though it is more efficient than a grid.
Edited by gwgill - 2/20/14 at 2:50pm
Quote:
Originally Posted by Light Illusion

I don't understand this, but I just took the original LightSpace and Argyll LUTs as supplied by Zoyd, and applied them to a test image.

These are the results.

.....

I suspect the colorspace choice is tricking you. It's not at all clear what you are looking at - perhaps raw RGB processed through the 3DLuts and then displayed in the RGB space of whatever screen the above raster images land on ?

[ And I'm not sure if there's much point closely examining two .jpeg images full of DCT artefacts. ]

You really can't make a visual evaluation of the device colorspace value without running it through the device itself to turn it from numbers into color.

While I will be the first to say that ArgyllCMS is not perfect, and I have a list of things I think can be improved, (unlike, say LS, which I gather is always perfect), the results are in the first posting - excellent delta E's and no visual difference to other profiling packages.
Edited by gwgill - 2/20/14 at 3:40pm
Quote:
Originally Posted by gwgill

It's not at all clear what you are looking at

Yes it is

[ Duplicate post deleted ]
Edited by gwgill - 2/20/14 at 3:39pm
Quote:
Originally Posted by buzzard767

Quote:
Originally Posted by gwgill

It's not at all clear what you are looking at

Yes it is

....

Well, I'm only guessing because I don't have the two 3DLuts in front of me, but I would suspect that the images include the full Video encoded range, and the difference is that the ArgyllCMS cLUT is careful to pass through the reserved sync levels (0 and 255 in 8 bit), whereas the LS cLUT is not preserving them.
Quote:
Originally Posted by buzzard767

Yes it is

I just checked the vertical color ramp from Ted's disk on the actual device the two LUTs were intended for and both LS and Argyll LUTs look smooth in that area without that edging effect you have highlighted. If you'd like I'll post screen shots but I don't think they are particularly informative.
Edited by zoyd - 2/21/14 at 2:05pm
Quote:
Originally Posted by Light Illusion

Display devices are not perceptual, they treat all colours the same - and as it is the display we are calibrating, not the uses perception of the display, that is what is important.

The bolded comment is exactly correct and why the whole idea of the CIE standard observer was developed in the first place, which eventually led to color references (and calibration to those references) that could be based on physical units. And you can measure how well you match the reference standards developed for the standard observer in many ways, the least useful of which is how well you match the device output to your desired RGB values. This tells the user absolutely nothing about the quality of the calibration, thereby ensuring an open-ended process because you have no physically meaningful goal. For the same reason that probe sensitivities are specified in some type of physical unit to distinguish it's value as a tool compared to another probe, so too are perceptual metrics used to assess the quality of the calibration with respect to the standards.
Quote:
Originally Posted by TomHuffman

The only way anyone can have a public discussion about this stuff is to use objectively communicable data. Simply asserting that a result is "better performing" because. . . . why? is not informative to anyone other than the speaker.

Nicely put, if you can't express your point regarding relative performance of this or that calibration detail without either posting data or referencing someone else that's done the work, you are just adding noise to the conversation.
Edited by zoyd - 2/20/14 at 4:57pm
Quote:
Originally Posted by zoyd

[snip]

Nicely put, if you can't express your point regarding relative performance of this or that calibration detail without either posting data or referencing someone else that's done the work, you are just adding noise to the conversation.

In that context here is the gamut of my 65VT60 with just the internal CMS. Using a LUT would seem to be a waste of money and time. I didn't post the color checker data for my 60VT65 since I felt no need to run that test on it.

Larry
Quote:
Originally Posted by LarryInRI

In that context here is the gamut of my 65VT60 with just the internal CMS. Using a LUT would seem to be a waste of money and time. I didn't post the color checker data for my 60VT65 since I felt no need to run that test on it.

Larry

Larry, from previous posting, I know you are using 65 Contrast and power gamma 2.2. Are you also using WARM1 and REC 709, and also setting GAMMA in Panasonic to 2.2 (or 2.4)?
I also try just using internal CMS and got good result, but not as good as you on the color checker. More specifically, a few color stand point (dE ~5).
Hey guys, anyone who believes his display is perfect and the 3D LUT is a waste of time, he can download the LightSpace DPS and measure his perfect display with 17-Point Cube and later post us the 3D Cube Viewer Presentation, RGB Separation Charts etc.

If he needs more detailed report, ask from Steve a 1-Day Trial of LightSpace and export that measurement file, I will enter this to DisplayCalibrationTools to report to you the dE you have to all measured points.

These reply's of perfect internal control displays are really very funny.

At 2012 when there were no tools in HT world for Large Cube Profilings, all they were believing that with 11-Point Grayscale and 6 Gamut measurements only; that defines a reference display..... later came the Lumagen with 5x5x5 feature and all who tryed it show greater picture from their 'perfect' displays.

From 11-Point Grayscale with 6 Color Gamut -> 5-Point Cube -> 9-Point -> 17-Point -> 21-Point.... yes all steps have a difference, but you have to see it with your eyes to believe it, charts can be good to all these steps.

All people here with 3D Cube LUTed displays, they have seen all these Cube Sizes differencies to their displays can understand the difference.

Also there 3 LightSpace users I know with xxVT60 with eeColor / Lumagen who knows well the 3D-LUT difference to their displays.
Edited by ConnecTEDDD - 2/21/14 at 5:39am
An honest serious question, if the dE's tell us that there is no significant difference, how do we know the improvements you see are not completely subjective? And basically the same as someone hearing more "black between the notes" because they've put wooden blocks underneath their speakercable?
Quote:
Originally Posted by Wouter73

An honest serious question, if the dE's tell us that there is no significant difference, how do we know the improvements you see are not completely subjective? And basically the same as someone hearing more "black between the notes" because they've put wooden blocks underneath their speakercable?

If dEs don't change after a 3dLUT then nothing has improved. But you better be measuring dEs across the board in order to verify whether the dEs have indeed not changed.
Quote:
Originally Posted by Wouter73

An honest serious question, if the dE's tell us that there is no significant difference, how do we know the improvements you see are not completely subjective? And basically the same as someone hearing more "black between the notes" because they've put wooden blocks underneath their speakercable?
Just because dE was developed for CIE standard observer (average human perception). But not all people are the same. Moreover, some have superhuman perception. And some have technologies far ahead of public science.

All we can do is to try it all yourself or blindly believe someone.
Quote:
Originally Posted by Vitalii427

Just because dE was developed for CIE standard observer (average human perception). But not all people are the same. Moreover, some have superhuman perception. And some have technologies far ahead of public science.

All we can do is to try it all yourself or blindly believe someone.
Quote:
Originally Posted by Vitalii427

Just because dE was developed for CIE standard observer (average human perception). But not all people are the same. Moreover, some have superhuman perception. And some have technologies far ahead of public science.

All we can do is to try it all yourself or blindly believe someone.

Agree with this.

In fact I seem to remember a very interesting program on a BBC Horizon Documentary with a subject matter of colour.

http://www.bbc.co.uk/programmes/b013c8tb

One aspect was the view that the way how we perceive colour depends on where we live, in other words we learn what to differentiate between..
It demonstrated that people in one part of the world were (I think this is correct) unable to see the difference between blue and green. Yet they saw massive differences between areas that the rest of us were unable to perceive.

Not sure if this makes a case for or against patch use.
.
Edited by PE06MCG - 2/21/14 at 2:28am
It doesnt matter where the purpose is calibration. How it is percieved is irrelevant, what matters is that the display is set to the standard.

If in Europe a meter (100 cm) is considered very large and the same meter is considered extremely small in Japan doesnt matter. The meter still has to represent the same distance.

There are more variations coming as we are in the process of finalising a program to automatically build optimised patch sequences based on user input requirements.
All our optimised patch sets are based on approaching all colours equally.

There two initial profiles, One that is equivalent to a 21^3 profile, but with just 2,330 patches rather than 9,261, and a second that has 9,008 patches, and is equivalent to a 33^3 cube profile, or 35,937 patches.

Steve
Quote:
Originally Posted by Light Illusion

There are more variations coming as we are in the process of finalising a program to automatically build optimised patch sequences based on user input requirements.
All our optimised patch sets are based on approaching all colours equally.

There two initial profiles, One that is equivalent to a 21^3 profile, but with just 2,330 patches rather than 9,261, and a second that has 9,008 patches, and is equivalent to a 33^3 cube profile, or 35,937 patches.

Steve
Steve, can you get us some measurable test results(numbers) of these sets? Either dE or what you call volumetric accuracy. How you compared these sets vs large grids?
Quote:
Originally Posted by Light Illusion

There are more variations coming as we are in the process of finalising a program to automatically build optimised patch sequences based on user input requirements.
All our optimised patch sets are based on approaching all colours equally.

There two initial profiles, One that is equivalent to a 21^3 profile, but with just 2,330 patches rather than 9,261, and a second that has 9,008 patches, and is equivalent to a 33^3 cube profile, or 35,937 patches.

Steve

great news Steve. Will try the new sequence as soon as my new toy arrives. Will You release this patch sequence program to Your users?
Quote:
Originally Posted by Wouter73

It doesnt matter where the purpose is calibration. How it is percieved is irrelevant, what matters is that the display is set to the standard.

If in Europe a meter (100 cm) is considered very large and the same meter is considered extremely small in Japan doesnt matter. The meter still has to represent the same distance.

Not disagreeing with the objective, simply making the point that the perception of your hardware accuracy is likely to be different depending on many factors.

Tuning in to certain areas of 'known' human extra perception may not be as universal as you may think. Maybe it needs to be selected bearing in mind several other factors ?
Perhaps this perception depends on their place of birth, language or learning during this growing up period.?

Your example of Japan is also interesting. I note that Calman includes a 9300K base for calibration. Will this have an effect on perception?

Sorry, just a novice with an interest in calibration.
Quote:
Originally Posted by Wouter73

An honest serious question, if the dE's tell us that there is no significant difference, how do we know the improvements you see are not completely subjective? And basically the same as someone hearing more "black between the notes" because they've put wooden blocks underneath their speakercable?

I figure it'll be months maybe a year when the dust settles and the truth be known.

I'd be more interested in trying it myself if they weren't still trying to bang out all the details. Really, shouldn't this pattern stuff been figure out months ago?

Not that keen on the complexity of implementation nor the cost either.

I suspect that on the whole, Tom is right. Some display don't benefit enough to go cube. Others only need a fewer number of correction points. When DO you say that you've reached the point of diminished return?
Edited by JimP - 2/21/14 at 3:48am
The problem I have is that the results that are claimed are not measurable. It is said that an even larger LUT gives inexpected better results, but when asked for proof the answer is "you have to see it". Ofcourse this could be completely true. But for me to purchase an ee colorbox and LS, for quite a large amount of money, I need more.

There is a mechanic that COULD be working here. You buy equipment for a lot of money, invest a lot of time. You WANT it to work better. So when you judge the result, it seems to work better. Not saying that this is the case here, but without anything measurable, it could be... Believe me, if I had the funds to buy things like this without a care, I would have already .
One point that I was not clear on was the results of using a color gamut that was closer to Rec 709 to start with. Instead, used native which needed more correction.

Seems that if you start off closer to your target, then the non corrected points wouldn't be as far off.
Quote:
Originally Posted by DrFaxe

great news Steve. Will try the new sequence as soon as my new toy arrives. Will You release this patch sequence program to Your users?
It's all work-in-progress at the moment, but we may indeed release the program to registered users - or integrate it into LightSpace. Not sure yet.
The cose and the patch sequences generated are not yet fully optimised...

We're simultaneously working on other improvements, so will decide when all is finalised.
Quote:
Originally Posted by Wouter73

The problem I have is that the results that are claimed are not measurable. It is said that an even larger LUT gives inexpected better results, but when asked for proof the answer is "you have to see it". Ofcourse this could be completely true. But for me to purchase an ee colorbox and LS, for quite a large amount of money, I need more.

There is a mechanic that COULD be working here. You buy equipment for a lot of money, invest a lot of time. You WANT it to work better. So when you judge the result, it seems to work better. Not saying that this is the case here, but without anything measurable, it could be... Believe me, if I had the funds to buy things like this without a care, I would have already .

Which is why it is so important to be able to quantify the limitations of the system in commonly accepted physical units. That way everyone has the same "value system" with which to make informed decisions and not have to rely on someone else's promise that the grass is greener over here and possibly waste their time and money.
Quote:
Originally Posted by Vitalii427

Steve, can you get us some measurable test results(numbers) of these sets? Either dE or what you call volumetric accuracy. How you compared these sets vs large grids?
The problem is that were are now getting to the point where standard measurement metrics just don't work...

The only real way is to measure as many individual colour points as possible throughout the volumetric space, and then assess the delta-e for each and every one.
That means using a very large profiling sequence, and using something like http://displaycalibrationtools.com/ to report on each point measured.

It would be an interesting evaluation to perform, but rather time consuming

(I use the teem delta-e just because it's a known concept, but what you need is to know the deviation from the perfect actual value)
Quote:
Originally Posted by JimP

One point that I was not clear on was the results of using a color gamut that was closer to Rec 709 to start with. Instead, used native which needed more correction.

Seems that if you start off closer to your target, then the non corrected points wouldn't be as far off.

That's typically the case, you can reduce the number of patches needed to meet a specific dE criteria if you start with a more linear color space (and vice versa).
New Posts  All Forums:Forum Nav:
Return Home
Back to Forum: Display Calibration
• A comparison of 3DLUT solutions for the eeColor box

### AVS Top Picks

AVS › AVS Forum › Display Devices › Display Calibration › A comparison of 3DLUT solutions for the eeColor box