The hue and saturation of color can be measured by the xy points on a CIE chart. However, since this chart does not reflect the intensity of color (Y), what"s the best way to measure this parameter?

Here are the specs for SMPTE-C intensity

Red 0.212
Green 0.701
Blue 0.087
Cyan 0.394
Yellow 0.457
Magenta 0.150
White 1

It looks like I could simply measure the light output of a white reference (75 IRE, 100 IRE?) and then compare the output of the various color fields as a percentage of white. However, I've tried this and I don't get what seems like correct results. In particular, I get very different results for the primaries and the secondaries.

Can someone steer me in the right direction on this?

By Eye. :)

This is one parameter that using disks for calibration wont help.

You would need an extremely accurate light meter that had equal response across the spectrum in order to measure this accurately. Your average colorimeter just won't have this kind of accuracy.

I would suppose that if your greyscale is correct, you should have these figures anyway.

Secondary intensity should be additive of the two primaries, as white intensity should be additive of all the primaries - what is your source for your numbers?

If your numbers don't add up then try to get more light into your sensor (smaller image with PJ and direct lens reading)

WideScreen Review has printed it a few times - you can also get from the RGB to XYZ matrix ( Poyntons REC709 http://www.poynton.com/notes/colour_and_gamma/ColorFAQ.html#RTFToC18) - then just make the secondaries additive. Notice the the middle row of the matrix adds up to Y=1 if RGB=1 so these are proper normalized numbers.

Looking closer your secondary numbers are the average of the primaries - that is not right - it should be 2x that - you got those numbers from ColorFacts right?

Kraz:

If the secondaries are supposed to be the sum of the primaries rather than the average, then that solves the problem. I was getting a correct result for the primaries, but the secondaries seemed way too low, by about half.

Mystery solved. Thanks.

BTW, if memory serves I got the numbers from ColorFacts.

Thus, I ought to be able to use a light meter, measure the output of a white field, and then calculate the proper intensities of RGBCYM as percentages of white. Correct? So, as a percentage of white, I should measure:

Red 21.2%
Green 70.1%
Blue 8.7%
Cyan 78.8%
Yellow 91.3%
Magenta 29.9%

Yes?

Kraz:

If the secondaries are supposed to be the sum of the primaries rather than the average, then that solves the problem. I was getting a correct result for the primaries, but the secondaries seemed way too low, by about half.

Mystery solved. Thanks.

BTW, if memory serves I got the numbers from ColorFacts.

Thus, I ought to be able to use a light meter, measure the output of a white field, and then calculate the proper intensities of RGBCYM as percentages of white. Correct? So, as a percentage of white, I should measure:

Red 21.2%
Green 70.1%
Blue 8.7%
Cyan 78.8%
Yellow 91.3%
Magenta 29.9%

Yes?
Tom - Those are pretty close to what we get for SMPTE-C. If you set the Primary Levels in CalMAN to 100%, then you can check these against the Yn reference for the primaries and secondaries. If you use a different value for Primary Levels, then the proportion is the same, but we correct for gamma.

Bill

I don't have my SMPTE-C numbers handy - the other computer upstairs must be powered off as I cannot browse the network. But adding is the right thing to do. Once you start measuring you will find this measure is the one that most manufacturers marketing games are played with - and no adjustment capability - and it is a measure that is never discussed in reviews!

I absolutely agree with your other thread about color being calibrations dirty secret- either the video decoding or display primaries muck up the numbers. I have been measuring xyY of color as well as greyscale for a while now (I prefer CIELCH or CIELUV for presentation though)

But adding is correct - indeed CF threw me for a loop when I first used it as a number source - but since they no longer have a bug tracking forum - it has never been fixed!

OK, one last complication. We have the correct proportions. Now the question is proportions to what? For example, on the Avia Pro Color Field menu, there's a selection for both White and Gray fields.

I would think that white would be the correct reference, but that yields color intensities that are WAY too high. So I use the gray field as my reference? Why gray instead of white?

Do you mean 100% vs. 75% patterns? The brightness ratios should remain the same regardless of the pattern level as long as you are not mixing up your pattern levels (of course things like BrilliantColor or WhitePeaking on DLP or DynamicIris/Lamp on LCD/SXRD means it will not be because it varies brightness based on content).

Which menu path are you using in AVIA PRO - so I can look and see. Normally I use my Accupel for these measures.

Color, Color Fields.

OK, one last complication. We have the correct proportions. Now the question is proportions to what? For example, on the Avia Pro Color Field menu, there's a selection for both White and Gray fields.

I would think that white would be the correct reference, but that yields color intensities that are WAY too high. So I use the gray field as my reference? Why gray instead of white?
These are proportional to white/gray. If you use a 75% white pattern, which would be ~53% for a 2.2 gamma, then the primaries and secondaries will have luminance/illuminance values based upon this. In other words, a 75% blue primary has a luminance (Y) value of around 4.5%. There is a reason why measuring blue is somewhat of a challenge! :)

Bill

Are these proportions valid for Rec. 601? Rec. 709?
Thanks.

Red 21.2%
Green 70.1%
Blue 8.7%
Cyan 78.8%
Yellow 91.3%
Magenta 29.9%

Yes?


No. :p

Are these proportions valid for Rec. 601? Rec. 709?This is for Rec. 601. The Rec. 709 proportions are a little different.
Red: 21.3%
Green: 71.5%
Blue: 7.2%
Cyan: 78.7%
Yellow: 92.8%
Magenta: 28.5%

No. :pIs there something constructive you'd like to add to this discussion?

Yes, the numbers you came up with i just dont think you can accurately come to the conclusion that those are ideal settings. I say learn to know what type of picture you like. Are you watching to make someone else happy or you? I've seen to many displays where people say they used service menu tweaks and every calibration cd in the book, yet the picture looks horrid, but they believe it's accurate.

Zues you need to understand that both TomHuffman and krasmuzik are pro calibrators and have the equipment and software needed to be.

I think Tom has a very good question that no one has asked before.

I didn't exactly "come up" with these numbers. These figures are the official specifications for HD and SD color spaces.

Yes, the numbers you came up with i just dont think you can accurately come to the conclusion that those are ideal settings
They are the defined standards. The primaries and secondaries in the standards are comprised of both color location (e.g., xy coordinates) and a relative luminance -- what is being discussed here.

I say learn to know what type of picture you like. Are you watching to make someone else happy or you? I've seen to many displays where people say they used service menu tweaks and every calibration cd in the book, yet the picture looks horrid, but they believe it's accurate.
Please take Derek's advice to heart. Service menu plus "CD" sans measurement equipment and knowledge equals an unhappy mix.

Bill

I think Tom has a very good question that no one has asked before.


And i dont think know one will answer it like me before. :D Color intensity, seperates the men from the boys in being able to determine 'too hot' unaccurate. Most of the time i would say the colors are not to bright, just the display cant accurately display the hot colors, which i bet most "pro" calibrations would look like to me. A pro calibrator that dont use "judgement" from his own eyes, the best tool for VIDEO, is not a pro imo. There is nothing that makes his calibration different from someone elses, their is no "signature" look to his calibration.

You would need an extremely accurate light meter that had equal response across the spectrum in order to measure this accurately. Your average colorimeter just won't have this kind of accuracy.

I would suppose that if your greyscale is correct, you should have these figures anyway.
Not necessarily. If your grayscale is dead-on accurate and your primaries are dead-on accurate, then you will get this mix exactly if you adjust color and tint to make your secondaries dead-on accurate. However, few displays have accurate primaries, so you generally have to change color and tint. Changing the color decoder controls essentially tweaks this mix in various ways for the secondary colors.

Hue ought to adjust the relative mix of intensities between the two additive primaries that make a secondary, and saturation ought to control how much of the "off" primary is added into the secondary to, wait for it, desaturate that color. Reality, of course, varies.

Bill

But what effect do primaries have on D65K? Does that not *have* to have a specific mix of R, G, and B?

But what effect do primaries have on D65K? Does that not *have* to have a specific mix of R, G, and B?
Yes, for a given set of primary locations, there will only be one mix that gets you to precisely D65. However, the primaries do not necessarily, in and of themselves, actually define where the secondaries are (they should be straight addition, but you can cheat a bit [see below and above]. For digital inputs, some displays will assume something like straight addition (aka the above mix) to get the secondaries. This is where your secondaries get placed on a line between the opposing primary and the white point.

Alternatively, you can adjust this mix just like you do in the color decoder so that you can get accurate secondaries even if the primaries are inaccurate. The question is also where this adjustment occurs: when decoding the inbound signal or potentially in the imager/formatter itself (or some set of intervening electronics). Chances are that if the display allows for color/tint changes when fed digital RGB data, it is doing this latter function.

The net point of all of this is that you can calculate a target luminance level for each primary and secondary, just as you can with white. This allows you to calculate color error (dE) for each point, and a goal for adjusting the controls that affect the points on the rim/edge of the gamut will be to minimize that color error if your goal is total accuracy in both the grayscale and in the rest of the gamut.

Bill

A pro calibrator that dont use "judgement" from his own eyes, the best tool for VIDEO, is not a pro imo. There is nothing that makes his calibration different from someone elses, their is no "signature" look to his calibration.

A professional calibrations SHOULDN'T have a "signature" look. The whole point is to acheive an accurate reproduction of the source. If calibrators are doing what they should be doing you shouldn't be able to tell one calibrator's work from another's (excepting additional services like lens striping and duvetyne linings).

The only calibration signature possible is the art of compromise that comes when the displays controls do not allow it to be perfected on all measures. A true artist would use psychovisual science when making those decisions.

Is there something constructive you'd like to add to this discussion?

Don't mind him, he is skilled in the ways of trollage.

You would need an extremely accurate light meter that had equal response across the spectrum in order to measure this accurately. Your average colorimeter just won't have this kind of accuracy.
What is accurate enough colorimeter to adjust (map) primaries to the close enough location in xyY space for LCOS/UHP PJ like Pearl reading from the lens? Is LT enough or one needs an Eye-One pro or something better?


I would suppose that if your greyscale is correct, you should have these figures anyway.Try to play with saturation control and measure "these figures" :)

A professional calibrations SHOULDN'T have a "signature" look. The whole point is to acheive an accurate reproduction of the source. If calibrators are doing what they should be doing you shouldn't be able to tell one calibrator's work from another's (excepting additional services like lens striping and duvetyne linings).

The point is accuracy, but I have not seen a display that is completely accurate. The differences in "look" are going to come down to the accuracy of the tools employed and the skill of the person using them. Here is a link to a customer of mine who found one person's final product vastly different from four others he used. A trained eye is key when it comes to making judement calls on what to give a little here or there on a specific sample.

http://www.avsforum.com/avs-vb/showthread.php?p=10082255#post10082255