Talk:Relative luminance

Origins of the term luma

Some of the information here seems to be a misinterpretation of Poynton's information on the subject?

The terminology luma was not introduced by the NTSC in 1953... they used the term "quantity representative of luminance".
Charles Poynton (according to him anyways) campaigned for the usage of the term luma.

An incorrect luma setting can cause visible flicker in a television display.

I don't believe this is true.... how can you have an incorrect luma setting?

Some nitpicks and inaccuracies/confusions

For standard-definition television systems using older phosphor characteristics modeling for the RGB primaries, (non-linear) luma can be calculated from gamma-adjusted RGB using the CCIR 601 formula Y' = 0.299 R' + 0.587 G' + 0.114 B'. Modern HDTV systems use a different definition of RGB phosphor characteristics and the ITU-R BT.709 formula Y' = 0.2126 R' + 0.7152 G' + 0.0722 B'. In both formulas, scaling and offsets used on both sides of the equation are assumed to be the same (e.g., ranging from 16 for black to 235 for white when used for digital video such as CCIR 601). Otherwise, some adjustments to the formula are needed to account for different scaling and offsets.

1- The CCIR was superceded by the ITU-R... so one could just stick to saying ITU-R Rec. 601 and ITU-R Rec. 709 Not all HDTV systems use the Rec. 709 primaries (i.e. phosphor characteristics / chromaticities) and luma co-efficients... see Charles Poynton's book Digital Video and HDTV for more information.

2- The information on scaling and offsets doesn't seem to be quite relevant. The R' G' and B' components don't have scaling or offsets as far as I know (other than camera settings), as suggested by the phrase "on both sides". Regardless, it's kind of bizarre to refer to those components as having scaling or offsets.

3- RGB should have prime symbols on it (when referring to gamma-corrected components, which is the case for luma).

4- The phrase "(non-linear)" seems to be unnecessary, as it was just explained and can only add to confusion. i.e. if there's a 'non-linear' luma, there's also a linear luma?

Confusing

This page should be merged into luma (video), with luminance (video) redirecting into luma (video).

Also, some of the information isn't as clear or accurate as it could be? Charles Poynton's book has some good information once you understand it, although it is not easy reading!! also see poynton.com Glennchan 06:09, 16 November 2006 (UTC)[reply]

Uh, I guess you noticed how confused I was. You're right. I don't see why it wouldn't be smarter to cover both words in one article, so we could keep them in sync and distinguished appropriatedly. Dicklyon 06:49, 16 November 2006 (UTC)[reply]

Hold on here. :D Television and video pretty much always use LUMA (gamma correction first, then forming the weighed sum), not "luminance" (forming the weighed sum, then gamma correction). See the chroma subsampling article; hopefully that explains it. Then read poynton's article on this. http://poynton.com/papers/YUV_and_luminance_harmful.html Glennchan 04:37, 17 November 2006 (UTC)[reply]

Images showing the difference. In both images, the chroma is subsampled with the 4:1:1 scheme. It helps show the difference between luma and luminance. The first image has constant LUMA. The second image has constant LUMINANCE.

luma

luminance

I will try to give a better explanation and copyright-free images.

Hey Dicklyon, sorry about reverting the information but my research indicates that video uses luma- a weighted sum of gamma-corrected R'G'B' components. (And gamma-corrected implies non-linear components; the transfer function used to apply gamma correction is a non-linear function.) If you read Poynton ( [1] ) you can see a credible source explaining this.

You are correct that video uses gamma-compressed luma. And I agree that the title of this article was not so great. But why are we leaving an articles if it's not about anything? Maybe I'll move it luminance (color space) or something like that instead. Dicklyon 03:16, 18 November 2006 (UTC)[reply]

I think this page should really be a disambiguation page. The information in this revision seems to be a (confusing?) duplicate of the information in the luminance (physical photometric) page. Some of the information seems to be incorrect or confusing, because sRGB and Rec. 709 use non-linear components; the formula given is for linear components; in practical usage with sRGB and Rec. 709, you probably won't use that formula; if you do, you want the sister formula to convert from gamma-corrected components to linear components.

2- Could you please cite your sources? This should help clear up confusion (because people can read the original source and clarify the wikipedia article). It would also clear up inaccuracies. (It's also wikipedia policy.) Glennchan 21:09, 19 November 2006 (UTC)[reply]

Sorry about the confusion. I'll add some sources. The sRGB colorspace uses gamma nonlinearity as you note, but this article's topic is about colorimetric measurements in a linear colorspace using the same primaries. This is a key step in getting from a different color space such as an image sensor's space into sRGB. Dicklyon 21:15, 19 November 2006 (UTC)[reply]

LUMA should be separate (with a related link) from luminance, it is a different term and use. What follows is my attempt to remove some of the confusion I read above:

Gamma, Linearity, and related terminology is confusing because it is often misused. The authorities are Poynton, SMPTE, IEC, EBU, & NAB. However there are even discrepancies among these authorities on some minutiae.

One HUGE problem is the misuse of the term "linear." The "most correct" usage of linear implies "photometrically linear" which also means linear light or linearized colorspace. in THIS context, linear means how light works in the real world, which is simple math. I.e. If you have a light X that is radiating 100 photons onto a surface, and you add a light of the same intensity, you will then be radiating 200 photons to the surface - two lights, twice as many photons. ----BUT---- We as humans WILL NOT see it as a doubling. In fact we will perceive twice the photons as a very minor increase in light. Human vision is NOT linear. LINEAR is a STRAIGHT LINE, flat, NO gamma, and NO power curve. Linear is also represented as "Gamma 1.0" to clarify that it has no curve.

WHY is LUMINANCE? (LOL, sorry bad joke -- Luminance is Y) ---- Luminance is the Y in CIEXYZ, and it relates to LIGHT as it exists in the real world (it is light over an area). It is LINEAR as it light, it does not have a gamma or power curve, as a result it is NOT perceptually uniform. Unfortunately luminance and Y have been misused. Luminance and Y are linear like real world light. Most image storage and transmission however is usually encoded with some form of power curve or log curve or some combination thereof, meaning the data does NOT have a linear 1:1 relationship with real world light. It has a curve, not a straight line. Luminance should never be used to describe anything with a curve. Y' (Y prime) refers to the gamma curve in luma. In theory, Y (no prime symbol) should only refer to linear luminance... BUT commonly (groan) YUV, YIQ, YCbCr, all use Y but refer to gamma encoded luma, NOT luminance (even though the Y does not have the prime symbol).

TERMS:

luma or Y′ — gamma encoded level of brightness in a video scene without regard to color. Functionally the same as a black & white signal.

luminance or Y — linear light value (no gamma encoding) over an area. Luminance DOES NOT APPLY TO VIDEO ENCODING, though of course it can describe the luminance in the scene being photographed, but luminance should not be used to discuss the video signal.

Put another way, luminance is "scene referred" while luma is "display referred".

luminosity — light energy emitted over time. If there was to be a combination of pages/terms, luminance and luminosity might go together with a discussion of their differences and their relationship. This makes more sense than combining luma and luminance IMO.

Transfer Curve — refers to a gamma curve, log curve, compound curve, or LUT as a means to convert from one image value to another.

gamma curve — refers to a curve that uses a power function to define it (such as (linear image data)^0.45).

log curve — refers to a curve that is logarithmic in nature, as is film negative.

compound curve — more properly just stated as a transfer curve, such as sRGB which is part power curve and part linear at a tangent angle.

LUT — Look Up Table, essentially a bit map or other table of values to transform an image signal. LUTs can be used to apply any of the above curve types, and sometimes are in ICC profiles.

1D LUT — a simple LUT based transfer curve with only x/y dimensions of input to output values, regardless of hue/chromacity. (If I recall it's called 1D instead of 2D to differentiate from the x/y plot of a chromacity diagram.)

3D LUT — a "cube" LUT which transforms color taking hue/chromacity into account.

gamma compression: Not really a thing, use the term "gamma encoding." A gamma curve does not "compress" the image, in fact values 0 and 1 typically remain the same when a gamma curve is applied.

gamma non-linearity: somewhat redundant. "Gamma" implies a power curve, though "gamma curve" or "transfer curve" can be used to clarify if you're comparing to a "gamma 1.0" (flat line). Myndex (talk) 18:09, 10 April 2019 (UTC)[reply]

Lstar Wars: The Luminosity Menace

TLDR: Luma should not be combined with luminance. If combined with anything probably only with pages relating to video encoding. If anything perhaps there should be a page to disambiguate the many similar sounding terms used in colorimetry, and often misused, such as Luminance vs luminosity which both discuss light, but are two distinctly different ways of measuring light (i.e. of an area vs over time). I wrote a tongue in cheek comparison pf terms which tries to hi-light the issue (pun definitely intended)...

But that brings up why I'm commenting today: Wikipedia (including this article) is using the term "Luminosity Function" and while it is true this term was used by the CIE in the 20's & 30's, today the CIE uses luminous efficiency function, which is more accurate and less likely to cause confusion with the unrelated term luminosity which is light over time. "Luminosity Function" is not light over time, it is perceived strength per wavelength, and luminous efficiency function is a better descriptor. --Myndex (talk) 08:10, 10 December 2020 (UTC)[reply]

Rename page and change scope to relative luminance?

Ok I think it really makes more sense to talk about *relative* luminance. Color science both uses the concept of relative luminance (especially in the XYZ color space) and sometimes uses the concept of absolute luminance (at least I think so). The concept of absolute luminance would come up in things like the transition between mesopic and scotopic and photopic vision.

I don't know, the current page is a mess and makes little sense.

Suggested text:

Relative luminance follows the photometric definition of luminance, but with the values normalized to 1 or 100 with respect to a reference white. Like the photometric definition, it is related to the luminous flux density in a particular direction, weighted by the luminosity function (Y) of the CIE Standard Observer. The use of relative values is useful in systems where absolute reproduction is impractical. For example, in prepress for print media, the absolute luminance of light reflecting off the print depends on the illumination and therefore absolute reproduction cannot be assured.

plus add explanation of how we tend to see things in ratios, and not absolute values. Possibly add disambiguation information from luma (video), plus formulas to calculate luminance for sRGB color space. Add links to things like Lab color space.

Glennchan 05:12, 22 November 2006 (UTC)[reply]

Glenn, it's a good thing you're up on this stuff. I checked some books, and I'd say you're right that the "Y" that I was calling luminance is more often called relative luminance. So that would be a good name for the article. I'd be careful getting into not-so-accurate things like characterization of psychophysics via Fechner's law ("we tend to see things in ratios"); there's no need to introduce that here, or if you do then you have to talk about how it breaks down, and how Stevens's power law is a better description and is used to motivate L* etc. I'll be away for a few days, so have at it. Oh, and relative luminance values are often normalized from 0 to 1, as opposed to 100. Dicklyon 06:24, 22 November 2006 (UTC)[reply]

I don't read that many color science books so I don't know if the term relative luminance is used more than luminance. Although luminance would, more often than not, refer to relative luminance rather than the photometric or video signal uses of the word luminance. I don't know how to rename articles, so I'll be lazy and leave that to someone else. ;) I'm not very familiar with Fechner's or Steven's power law; what I was getting at is that our vision is more ?interested? about relative values than absolute values. Not talking about Fechner's or Steven's power law.Glennchan 06:00, 24 November 2006 (UTC)[reply]

You can use the "move" button to rename a page; it's easy, and makes the old name a redirect. You can consult lots of books at http://books.google.com. And if you're not familiar with Fechner's law and how it breaks down, stay away from that perception topic, and it's very easy to say things that are not right. Dicklyon 20:02, 26 November 2006 (UTC)[reply]

Ok, done. In regards to normalizing to 100, this is done in Adobe Photoshop and what is cited in Poynton's glossary for his book. I don't know what the practice is in the color science world, whether they use 100 or frown upon it. In a practical sense, it doesn't really matter too much since the number you normalize luminance to is kind of arbitrary.Glennchan 22:13, 29 November 2006 (UTC)[reply]

Information removed because it is extremely confusing (and some of it, not entirely accurate)

Relative luminance in colorimetric spaces

The deleted information:

For linear color spaces such as XYZ, Yuv, xyY, etc. (but not the nonlinear color spaces YUV, YCbCr, YIQ, L*a*b*, etc.), the Y or luminance component is explicitly a relative luminance. For nonlinear colorspaces such as YUV, where Y represents "luma", conversion to relative luminance requires first a reconstruction of nonlinear R, G, and B values, then a linearization, then a weighted sum to compute relative luminance. For L*a*b* space, the L* component is simply a nonlinear version of the relative luminance Y.

For colorspaces such as the sRGB color space and HDTV, and their YUV and other video derivatives, that use the ITU-R BT.709-3 primaries, the weights to compute luminance Y from linearized R, G, and B values can be found as the center row of the RGB to XYZ transformation matrix:^[1]

Y = 0.2126 R + 0.7152 G + 0.0722 B

Although colorimetric relative luminance is often used in connection with video and computer systems, it should not be confused with luma, the weighted sum of the nonlinearly gamma compressed RGB components.

Hey Dicklyon, I honestly am not sure what you are trying to get at with the information you added / was originally there.

HDTV does not always use the same primary chromaticities; hence, the co-efficients to calculate luminance are different. SMPTE 240M goes hand in hand with SMPTE C primaries; see the luma (video) article and grab the luma co-efficients, they correspond to the luminance co-efficients.
Y'UV refers to an analog video interface, with gamma-corrected components. As does Y'IQ. I'm not sure why you mention them.
"Nonlinearly gamma compressed" is a bizarre phrase. Gamma compression implies non-linearity. "Non-linear, gamma compressed" sort of makes more sense, although you don't need the word non-linear in there. I just stick it in there just to stress that the components are non-linear; it is redundant but intentionally so.
YCbCr should really be referred to as Y'CbCr. No one uses YCbCr (as in, a YCbCr color space based on luminance and not luma).
I don't understand what you're trying to get at.
If you want to show an example calculation to get relative luminance, the Lab color space article is good.
The use of the words linear and nonlinear can be confusing. Both Y' (from Y'CbCr color space) and L* are non-linear. However, there's a significant difference between the two. L* is calculated from linear components, Y' is not.

Glennchan 23:17, 29 November 2006 (UTC)[reply]

Sounds mostly right to me. So let's fix it. Dicklyon 00:09, 30 November 2006 (UTC)[reply]

I still have no idea what you're trying to get at. Are you trying to give an example where relative luminance is used? If so, just link to Lab color space pages (or give the same example). The mention of luminance for Y'CbCr color space is *extremely* confusing, since (A) Y'CbCr doesn't use luminance to begin with. It uses luma. (B) If you ever had to calculate luminance, it would likely be photometric luminance. Glennchan 04:00, 30 November 2006 (UTC)[reply]

Colors in Y'CbCr mostly map 1-1 with colors in RGB or XYZ, don't they? No matter what space you choose, you have a relative colorimetric representation, relative to a reference white. So you can convert to find Y. For L* it's easy because it's 1:1 with Y. For Y'CbCr, it's more complicated to get to Y, but still straightforward. Dicklyon 04:35, 30 November 2006 (UTC)[reply]

Y'CbCr does not map 1-1 to colors in RGB or XYZ. The color gamut is differently sized than RGB. Rec. 709 Y'CbCr for example defines colors that do not fit into the sRGB gamut; it defines some impossible colors with negative components (obviously, negative light doesn't exist). Although I'm not really sure what you mean by 1-1 mapping, because that makes no sense. Obviously a value like 12 209 94 RGB isn't going to be the same as 12 209 94 XYZ or Y'CbCr (where 12 209 94 are arbitrary numbers). As far as being straightforward, that's subjective and I'm not sure what you mean by that.

What I meant by "mostly map 1-1" is that in those regions where the gamuts overlap, which is "most" colors, there is a 1-1 mapping between the spaces. That does NOT mean the numbers are the same in different spaces, but that the colors represented are the same. And it does NOT apply 1-1 to the quantized 8-bit values, just to the underlying continuous values. Straightforward means just follow simple steps; it could be said differently if that's too subjective. Dicklyon 05:31, 30 November 2006 (UTC)[reply]

Mostly mapping 1-1: Couldn't you say that of ANY color space??? I don't follow. You need to be specific. Glennchan 06:00, 30 November 2006 (UTC)[reply]

Yes, sorry, that was the point. Any pair of colorspace. Dicklyon 06:07, 30 November 2006 (UTC)[reply]

2- What are you trying to do? What are you trying to say? What kind of information are you trying to put into this article? (that's not already there) Glennchan 06:00, 30 November 2006 (UTC)[reply]

How to compute relative luminance for an arbitrary color. Dicklyon 06:07, 30 November 2006 (UTC)[reply]

3- I think you are extremely confused. For example, it seems like you assume that color spaces (for computer and video imaging) use linear non-gamma corrected components, which is not the case for R'G'B' or Rec. 709 Y'CbCr. Perhaps it's things like that which are tripping you up? Glennchan 06:00, 30 November 2006 (UTC)[reply]

I'm not making any such assumption. Trying to provide answers that work for most kinds of colorspaces. However, be aware that whatever colorspace is used, there is usually assumed an underlying additive (linear) RGB color model related to CIE XYZ colorimetry. Since that's true for all R'G'B' and Y'CbCr spaces, one can compute Y by going through the right few simple steps, involving linear RGB on the way to Y. Dicklyon 06:07, 30 November 2006 (UTC)[reply]

Y'CbCr does have a relative representation; However, it does not use relative luminance. Yes you can convert to find Y; I think in some version of this article, it is stated that this is possible. You have to first convert the colors into linear RGB. To be more specific: apply a matrix to convert from Y'CbCr to R'G'B'; apply the inverse transfer function on each of R', G', and B'; then, apply the matrix to convert RGB into XYZ; or just take the middle row if you only want Y.Glennchan 05:20, 30 November 2006 (UTC)[reply]

Exactly. Isn't that about what I wrote? Dicklyon 05:31, 30 November 2006 (UTC)[reply]

References

^ Michael Stokes, Matthew Anderson, Srinivasan Chandrasekar, and Ricardo Motta, "A Standard Default Color Space for the Internet - sRGB", online see matrix at end of Part 2.

Relative luminance in colorimetric spaces Ping Pong

The text in question:

For colorspaces such as XYZ, xyY, etc. the letter Y refers to relative luminance. No computation is required to find relative luminance when it is explicit in a color representation in such spaces.

For RGB colorspaces that use the ITU-R BT.709 primaries (or sRGB, which defines the same primaries), relative luminance can be calculated from linear RGB components:^[1]

Y = 0.2126 R + 0.7152 G + 0.0722 B

For other sets of RGB primaries, different linear coefficients are needed to get relative luminance. In general, the coefficients are all positive, the green coefficient is largest and blue smallest, and the three form the middle row of the RGB-to-XYZ color transformation matrix.

For L*a*b* space, the L* component is simply a nonlinear version of the relative luminance Y, so Y is easy to compute as a function of L*.

Although luminance is occasionally used in video engineering (i.e. in measuring monitor brightness), it should not be confused with luma, the weighted sum of the nonlinear gamma compressed R'G'B' components. For colorspaces that use luma, such as Y'UV or Y'CbCr, where Y' represents luma, conversion to relative luminance requires first a matrix transformation to get nonlinear R', G', and B' values, then a linearization to R, G, and B, then a weighted sum to compute relative luminance.

Discussion

Hey Dicklyon, sorry to play ping pong here but I just removed the section in question since I don't understand what you're talking about. For example, you provide a link to R'G'B' color space (a gamma-corrected color space), while the formula is appropriate for the linear non-gamma corrected RGB space.

Please find a third party source that talks about luminance (relative) and paraphrase them. The sRGB article you provided doesn't really talk about luminance (relative). Glennchan 05:30, 30 November 2006 (UTC)[reply]

OK, I'll work on finding good refs. But why do you keep taking stuff out instead of working on improving and understanding it? The referenced sRGB article has the linear transformation at equation 1.8. Dicklyon 05:37, 30 November 2006 (UTC)[reply]

It's because I can't determine what you're trying to say. It seems like you are confused, and because you're confused I'm confused. I have no idea what you're talking about. And since I have no idea what you're talking about, I can't improve on what you are putting in. Glennchan 05:40, 30 November 2006 (UTC)[reply]

OK, I'm confused. About what part of what I wrote you find confusing. Any point in a colorspace has a relative luminance, and it's often useful to be able to compute it. Does that help as a starting point? Dicklyon 05:54, 30 November 2006 (UTC)[reply]

The formula Y = 0.2126 R + 0.7152 G + 0.0722 B may be more appropriate in the luminance (photometry) article. Like, stick in a subsection for luminance in video or luminance in imaging systems. Glennchan 05:43, 30 November 2006 (UTC)[reply]

I thought we had been around that loop already. We settled on relative luminance because that formula gives a relative value, luminance compared to a reference white. It is dimensionless, unlike the photometric absolute luminance. Why is this article here if we're not going to use it to explain the notion of relative luminance in colorspaces, and how it can be computed? Dicklyon 05:54, 30 November 2006 (UTC)[reply]

Nevermind about that comment about moving that formula into the photometry article, I was thinking about something else. Glennchan 06:18, 30 November 2006 (UTC)[reply]

Sounds like you need to sleep on it. Me too. Dicklyon 06:40, 30 November 2006 (UTC)[reply]

2- I think the Lab color space article has information about computing (relative) luminance. You figure out the luminosity function Y, and divide by Y_n (where Y_n is the Y value of the white point). I think that's it (I haven't looked deeply into that myself).Glennchan 06:18, 30 November 2006 (UTC)[reply]

Yes, that's right, but the Y and Yn are themselves already relative, typically. That is, if you have Y on a scale of 0 to 100, then Yn is 100, and you use it to normalize to get relative luminance from 0 to 1. The subsequent equations then can assume a 0 to 1 range. Dicklyon 06:39, 30 November 2006 (UTC)[reply]

You could plug absolute values for Y and Yn into that equation. I think this article should describe a general method for finding the relative luminance. It doesn't make sense to jumble the article with references to L* and Y'CbCr; these should go into subsections.

Yes, it works also for absolute luminance, and turns it into a relative one; my point is that it works just as well for starting with another relative colorspace color specification. Feel free to move info into a better organization. To me it doesn't feel jumbled, but maybe I'm too close to it. See if you find a better way. Dicklyon 04:17, 2 December 2006 (UTC)[reply]

Some of this information may possibly be better in the luminosity function (Y) article. Once you discover Y, it's trivial to divide by the white point to get a relative luminance no? Glennchan 04:04, 2 December 2006 (UTC)[reply]

I don't see how it would be fit there, since the luminosity function is a well-define thing on its own. The luminosity function is used, at least implicitly, in finding relative luminance for an RGB color specification; perhaps should say more about that here? Oh, and finding absolute Y is not usually an option in colorspace problems. Dicklyon 04:17, 2 December 2006 (UTC)[reply]

If one simply normalizes Y to a number / let absolute Y = 1 or whatever, then I don't think it's a problem. No? Glennchan 04:36, 2 December 2006 (UTC)[reply]

Sorry, I don't follow. What potential problem are you asking about, and what do you mean by letting absolute Y = 1? That doesn't sound like an absolute luminance. Dicklyon 05:02, 2 December 2006 (UTC)[reply]

The "problem" I refer to is: Oh, and finding absolute Y is not usually an option in colorspace problems.

When you are trying to calculate relative luminance, it's almost the same thing as calculating absolute luminance. You'd just have to divide luminances to get relative luminance; when you aren't given an absolute luminance, just arbitrarily set it / normalize it to something. Glennchan 05:13, 3 December 2006 (UTC)[reply]

OK, my comment was with respect to your statement "Once you discover Y, it's trivial to divide by the white point to get a relative luminance, no?" Yes; but I pointed out that discovering Y is not an option usually, so all is relative. Sounds like you agree. Dicklyon 05:30, 3 December 2006 (UTC)[reply]

References

^ Michael Stokes, Matthew Anderson, Srinivasan Chandrasekar, and Ricardo Motta, "A Standard Default Color Space for the Internet - sRGB", online see matrix at end of Part 2.

Semantics - my bad

Regarding the reference of "RGB primaries" --> "display primaries" --> "primary chromaticities"
Primary chromaticities is a better phrase to use since it describes a concept more exactly. RGB primaries is sort of an ambiguous phrase, since primaries imply the primary colors red, green, and blue. Granted, the word primaries alone may suffice. Some people who use the phrase "primary chromaticities" would be [2] and Poynton's textbook. Glennchan 04:00, 2 December 2006 (UTC)[reply]

Yes, that's what I usually call them, too. It might be more clear to be more explicit and call them the red, green, and blue primary chromaticities. Up to you. Dicklyon 04:12, 2 December 2006 (UTC)[reply]

Y formula

The text:

For RGB colorspaces that use the ITU-R BT.709 primaries (or sRGB, which defines the same primaries), relative luminance can be calculated from linear RGB components:^[1]

Y = 0.2126 R + 0.7152 G + 0.0722 B

Why do you guys even need external reference when all formulas are there in sRGB? Why not just put it as

For RGB colorspaces that use the ITU-R BT.709 primaries (or sRGB, which defines the same primaries), relative luminance can be calculated from linear RGB components:

Y = 0.2126 R + 0.7152 G + 0.0722 B

? —The preceding unsigned comment was added by 195.137.203.137 (talk) 12:57, 15 January 2007 (UTC).[reply]

in fact I have just done so, feel free to roll back if you really have to 195.137.203.137

Wikipedia is never to be used as a source, because it is fluid and unreliable. Reliables sources should always be cited. Dicklyon 17:25, 15 January 2007 (UTC)[reply]

I am new to all this stuff and I am really confused. The formula is stated as:

Y = 0.2126 R + 0.7152 G + 0.0722 B

And yet the matrix on the "CIE 1931 color space" page implies a different formula:

Y = 0.17697 R + 0.81240 G + 0.01063 B

These values are completely different, even accounting for the scale factor - yet according to this article, the Y component of XYZ space represents relative luminance. [User:Anon] 10.55, 28 May 2007

The answer is that the matrix values Y = 0.17697 R + 0.81240 G + 0.01063 B are for CIERGB space, and Y = 0.2126 R + 0.7152 G + 0.0722 B is for Rec709 or sRGB space. There are different matrix values for every colorspace based on the primary and white point coordinates. Myndex (talk) 04:59, 9 October 2021 (UTC)[reply]

References

^ Michael Stokes, Matthew Anderson, Srinivasan Chandrasekar, and Ricardo Motta, "A Standard Default Color Space for the Internet - sRGB", online see matrix at end of Part 2.

[1] Michael Stokes, Matthew Anderson, Srinivasan Chandrasekar, and Ricardo Motta, "A Standard Default Color Space for the Internet - sRGB", online see matrix at end of Part 2.

[2] Michael Stokes, Matthew Anderson, Srinivasan Chandrasekar, and Ricardo Motta, "A Standard Default Color Space for the Internet - sRGB", online see matrix at end of Part 2.

[3] Michael Stokes, Matthew Anderson, Srinivasan Chandrasekar, and Ricardo Motta, "A Standard Default Color Space for the Internet - sRGB", online see matrix at end of Part 2.

[1]

[1]

[1]