Talk:CIELAB color space/Archive 1

Archive 1

This article needs reorganization (suggested split)

This article covers material which is broader than the title, currently "Lab color space", implies. Putting the Lab/L*a*b* (Hunter Lab vs CIELAB) issue aside, color spaces which do not even include "a" and "b" as part of the nomenclature are being mentioned and discussed, such as CIELUV. I could imagine someone adding the OSA color space, whose coordinates are L, j, and g, and possibly the CIECAM Brightness/Colorfulness rectangular appearance space with coordinates $Q$, $a_{m}$, and $b_{m}$. Is an article entitled "Lab Color Space" the right place for these, as well?

What I suggest is:

• Create an article for Opponent Color and Appearance Spaces. Mention that Photoshop and similar applications use (a possibly encoded version of) CIELAB, with a link (see next item). Topics common to all these spaces (rectangular to/from cylindrical, opponent mechanisms, Hering et alli, etc). Beside the Photoshop etc issue, which is what people seeing this mentioned in these programs will want to know first and foremost, it should be written from a colorspace-neutral perspective.
• Create individual articles for each opponent space: CIELAB, CIELUV, Hunter Lab, OSA, and CIECAM spaces. Maybe some historical ones, like U*V*W* and Hunter Lαβ, which would kind of clutter the existing works like the Hunter Lab thing seemed to have done.
• Remove the redirection from "CIELAB" to "Lab Color Space" as CIELAB will need a dedicated article of its own, and "Lab" ≠ "CIELAB," as somebody has pointed out. Likewise for "CIELUV".
• Redirect "Lab Color Space" to the new Opponent Color Space article mentioned above. Folks wondering what's under Photoshop's hood will need that discussion.

I can get some of these started, but I certainly don't want to mess around with the redirection business unless/until we're both ready and are in some kind of agreement. I believe one can edit a page which has a redirect at the top, and view it with an appropriate query string.

What say you, fellow editors and users? Shall this article be re-organized along these lines? Anybody have other (hopefully constructive) ideas? Lovibond 03:14, 5 April 2007 (UTC)

• I say go for it, in the spirit of WP:BOLD. --jacobolus (t) 11:00, 16 April 2007 (UTC)
• More specifically, I've been doing a bit of editing on the Munsell color system article, which is still woefully inadequate. It would be nice to somewhere in the wiki have a good historical explanation of color order systems, which can then link out to the Munsell article, this proposed article about opponent color spaces, RGB/CYMK color spaces, etc. I think all of your suggestions here are quite reasonable, so please don't hesitate to dive in. --jacobolus (t) 11:07, 16 April 2007 (UTC)
• I suggest you subsume the opponent process article in your new article about opponent color and appearance spaces. --jacobolus (t) 11:17, 16 April 2007 (UTC)
• [1] seems like a pretty reasonable description of Luv, Lab, and other modern color spaces. --jacobolus (t) 21:07, 16 April 2007 (UTC)

Thanks for adding "split" to the section header so I could it. I only half agree with the proposal. I agree that the Luv stuff doesn't belong here at all. And an article opponent color is a good idea. But I think we should leave Hunter Lab and CIELAB both in this article, since they are so closely related and need to be understood jointly to avoid confusions between them. Dicklyon 22:14, 5 May 2007 (UTC)

I agree with this somewhat. I think this article needs a much better discussion of the history and motivation of these spaces, and some diagrams, so that readers can get some intuitive understanding of what is being explained. And I think three of the main sections could be labeled “opponent color space,” “Hunter Lab color space,” and “CIELAB,” or similar, and detailed information about those can be split into separate articles (for instance the big blocks of formulae, which aren't of such interest to a general audience, I don't think), and then linked from those sections in the “Main article:«foo»” style. --jacobolus (t) 23:44, 5 May 2007 (UTC)

Okay, I just split the L*u*v* information into a new stub of an article, “CIE L*u*v* color space,” where it won't be cluttering up this page. So that goes some of the way. I'm actually not convinced that all the information about the Hunter Lab space belongs on this page. Does anyone still use that space anymore? Even hunterlab.com recommends using L*a*b*. It seems to me that Hunter Lab can be split into its own article, a note about the possible confusion between L*a*b* and Hunter Lab can be added in the introduction to this article (saying something like "the term Lab color space is ambiguous, and can also refer to the historical Hunter Lab space"), and then Hunter's work can be discussed in a history section on this page, as it led to the creation of L*a*b*. But the conversion formulae to and from Hunter Lab seem like mostly a distraction in the context of this page.

So my recommendation: make this article about L*a*b* exclusively, with Hunter Lab discussed in the intro/history sections, and then given its own article, where formulae, etc. can sit. I'd maybe change my mind if there's still any significant use or reason to use Hunter Lab today.

I would then maybe even suggest moving this page to CIE L*a*b* color space, which would make the title prettier and more correct. A large majority of people searching for Lab color are going to hit some sort of redirect anyway, so making the title accurate and pretty wouldn't have any downsides, I don't think.

--jacobolus (t) 23:33, 17 May 2007 (UTC)

Agreed. Lovibond 15:35, 29 May 2007 (UTC)

Okay, Lovibond, will you move the stuff you started putting at CIELAB into the history section of this article, and then maybe we can add a proposal to split out Hunter Lab, or Lαβ or whatever into other articles, linked out of sub-sections of the history section of this article, and a separate proposal to move this article to CIELAB color space, which would mirror the locations of CIELUV color space, RGB color spaces, etc. Dicklyon or others, do you have problems with changing this article to focus on CIELAB? I don't know of any current uses/users of Hunter Lab (I couldn't find any in web searches), so I'm not sure it's of more than historical interest, which makes the extensive emphasis given to it in this article a distraction.

I think the trickiest part of getting this article "right" is figuring out how to present some general information to newcomers, explaining why anyone would want such a color space, etc. without putting too much technical information that would scare them away, and providing lots of pictures. But also putting detail of the history, and technical aspects. I think expanding the "why use Lab" section and leaving it nearish to the top is not a bad idea, and then going on to history, and putting technical details at the bottom, then splitting out all the detailed Hunter Lab info into its own article.

Thoughts? --jacobolus (t) 23:57, 29 May 2007 (UTC)

I think focusing on CIELAB is great. The Hunter Lab stuff can be a separate article, or a minor subsection, but it shouldn't get in the way of the main topic. As far as explaining why anyone would want such a space, I'd say it's not so tricky; just find a source and summarize what they say. That's better than making up reasons for people's opinions or preferences. Dicklyon 00:17, 30 May 2007 (UTC)

Well, what is the main topic? The main topic of this article, based on its name, is "Lab color space", which is different from CIELAB. CIELAB is important enough that it deserves its own article, which is currently being written. Lovibond 15:48, 1 June 2007 (UTC)

In my opinion the main topic should be CIELAB, with other Lab-like spaces mentioned in the history section. --jacobolus (t) 22:56, 1 June 2007 (UTC)

So, Lovibond, are you willing to have this article be changed to focus on CIELAB, with Hunter Lab and other variants appearing in a history section but mostly discussed on other pages? I don't really think we need a "Lab color space" article at all, as CIELAB seems to be the only such space still in widespread use (even Hunter Lab recommends that CIELAB be used for most purposes). It seems like Dicklyon is in agreement with such a policy. I think your idea to “Create individual articles for each opponent space: CIELAB, CIELUV, Hunter Lab, OSA, and CIECAM spaces. Maybe some historical ones, like U*V*W* and Hunter Lαβ” is an excellent one. But I think “Lab color space” should redirect to wherever we put the “CIELAB” article, and a section near the beginning can link out to “opponent process,” and have a paragraph summary of the basic theory, that you suggest in your original comment under this header is necessary to understanding Photoshop's “Lab” color space. If you’re in (even general) agreement, then I think we’re good to go, and should get started. :) --jacobolus (t) 21:51, 10 August 2007 (UTC)

• From the above conversation, and lack of any more comment since 2007 I assume the split has been done, or is no longer required. I have removed the split tag as part of cleaning up the split backlog. If someone feels there is still a need for a split, please place a new split tag on the article with the current date, and start a new section on this talkpage with the current rationale. SilkTork *^YES! 10:54, 8 November 2010 (UTC)

Based on the above discussion, and due to the fact that CIELAB is in much more common use than Hunter Lab, I have renamed this article to "CIELAB color space" and focused the introduction on CIELAB, with the Hunter Lab material moved into the "Hunter Lab" section. (I think it is still a good idea to split out "Hunter Lab" into its own article, if anyone would like to take that on.) —Ka-Ping Yee (talk) 08:22, 8 May 2018 (UTC)

Ranges of a* and b*

Under the subsection Range of L*a*b* coordinates it states, "As mentioned previously, the L* coordinate ranges from 0 to 100. The possible range of a* and b* coordinates is independent of the color space that one is converting from, since the conversion below uses X and Y which come from RGB." That tells me nothing about the ranges of the a* and b* channels. What I'm really interested in is quantisation errors, specifically whether conversion from 16-bit RGB to Lab and back causes errors due to the (possibly small) number of discrete values of a* and b*. Perhaps I am missing something, perhaps either channel ranges from 0 to 1000000, but the article says very little about this. Ignoring colour, even an 8-bit per channel greyscale RGB image would seem to have more than double the number of tones than Lab (256 values versus 101). Surely this can't be the case? Please tell me I am wrong! nagualdesign (talk) 07:05, 24 November 2011 (UTC)

If 16bit RGB to Lab color is lossless conversion!?

In the article mentioned that: "... converting an RGB image to Lab and back was a lossy operation. With 16 bit/channel support now common, this is no longer such a problem." What does it mean? If it mean 16 bit RGB to Lab color is LOSSLESS? Further explanation is required in this sentence, in my opinion. — Preceding unsigned comment added by 91.99.58.81 (talk) 05:48, 29 April 2012 (UTC)

No, it just means that the loss due to quantization is negligible; "no longer such a problem"; not quite lossless though. Dicklyon (talk) 06:05, 29 April 2012 (UTC)

It seems your complaint is directed at the word "lossy". This word, however, is in the dictionary and a good understanding of it might (?) clear up any problems with the sentence.

Meaningless sentence

The following sentence does not make sense and needs to be corrected as appropriate:

These color appearance models, of which CIELAB, although not designed as ^[2] can be seen as a simple example,^[3] culminated with CIECAM02.

— 146.179.8.172 (talk) 16:54, 23 August 2012 (UTC)

Good catch. Looks like it was added by this anonymous edit in May 2009. –jacobolus (t) 20:39, 23 August 2012 (UTC)

Distance function

I have been led to believe that the Lab color space has been formulated in such a way that if one wants to determine whether color A is closer to color B or color C, then the ordinary Euclidean distance function is the best guide. In any case, a discussion of some kind of metric seems to be at the heart of any color space. — Preceding unsigned comment added by 71.56.157.70 (talk) 23:25, 7 September 2012 (UTC)

That was one of the most important goals of the CIELAB color space, yes. (Well, more precisely, the goal was to allow specification of color tolerances, for industrial purposes, such that colors matching some spec (of desired color and maximum color difference tolerance) would need to be within some reasonable standard of similarity, based on human vision.) Note, CIELAB was fairly successful, but the color difference metric it defines is far from perfect. [Also, a metric is not necessarily at the heart of every color space; the Munsell color system for example is not based around any Euclidean metric, though a few derived color difference functions have been proposed for it.] –jacobolus (t) 08:32, 2 April 2014 (UTC)

The CIE 2000 Delta E is based on CIE L*a*b* calculation is considered the best method for computing color difference (CIE Publ. 142-2001, CIE S 014-6/E:2013. While painful to calculate is gives a better estimate of Just Noticeable Difference (JND) and correlation to MacAdam's Ellipse sets based on human observations. — Preceding unsigned comment added by Jacarter3 (talk • contribs) 20:26, 17 October 2018 (UTC)

References

Which form of the Lab color space is used by Photoshop?

I can select several color modes in Photoshop, RGB, CMYK, Lab, etc. Is the Lab color mode in Photoshop actually Lab (Hunter Lab), or is it really L*a*b* (CIE Lab)?

Benhut1 (talk) 02:54, 25 April 2014 (UTC)

Photoshop has a CIELAB mode. No one has used Hunter Lab for anything in decades. –jacobolus (t) 02:47, 2 May 2014 (UTC)

An alternative undocumented LAB color space option exists in some applications called "LAB identity built-in". It might be helpful to clarify that as well. It is offered as an alternative to CIELAB D50 in Affinity, for example. 103.96.86.25 (talk) 12:19, 6 May 2022 (UTC)

Range of a* and b*?

In 2008, the following paragraph was added:

As mentioned previously, the L* coordinate ranges from 0 to 100. The possible range of a* and b* coordinates depends however on the color space that one is converting from. For example, when converting from sRGB, the a* coordinate range is [-0.86, 0.98], and the b* coordinate range is [-1.07, 0.94].

In 2011 this was changed to its opposite:

As mentioned previously, the L* coordinate ranges from 0 to 100. The possible range of a* and b* coordinates is independent of the color space that one is converting from, since the conversion below uses X and Y which come from RGB.

I find the latter sentence quite confusing. My question: What is the range of a* and b* in the CIE L*a*b* space? Surely when converting from some other color space to L*a*b*, the ranges of a* and b* may vary, because these other color spaces don't cover all colors. But the L*a*b* space itself must have some definite ranges for a* and b*, or not? AxelBoldt (talk) 00:34, 5 January 2016 (UTC)

In Photoshop and other software, a and b can vary in the range -128...127. But, as can be seen in the article's first image, which represents the body of ALL perceivable colors, the possible range is quite different, depending on the base color and the lightness which are just regarded. a and be can even go up to 160! (see mentioned image) Many values out of the CIELAB body can be produced by e.g. Photoshop, but they do not represent colors, they are nothing, they make no sense. --Farbenprofi (talk) 18:24, 26 February 2017 (UTC)

L*a*b* is based on a reflectance or transmission function and a reference or Standard Illuminant such as D65 with CIE XYZ coordinates at D65 = [ X Y Z ] = [ 95.0470 100.0000 108.8830 ]. Note that Y is chosen as 100. This represent 100% luminous flux (luminous intensity or power) and the coordinates for a reflective color surface must have Y <100. If the Y value of the color surface is close to 100 or to zero, then the range of a* and b* are very limited. if Y is between 30 and 50, then the range can be much larger. There is no fixed range for a* and b*. The equation to calculate a* is 500x the difference between two mostly cubic functions that range from zero to one; b* is 200x the difference between different but similar terms.Jacarter3 (talk) 20:03, 17 October 2018 (UTC)

In view of the above, shouldn't forward transformation be as such:

${\begin{aligned}L^{\star }&=\left(116\ f\!\left({\frac {Y}{Y_{\mathrm {n} }}}\right)-16\right)\times {\frac {255}{100}}\\a^{\star }&=500\left(f\!\left({\frac {X}{X_{\mathrm {n} }}}\right)-f\!\left({\frac {Y}{Y_{\mathrm {n} }}}\right)\right)+128\\b^{\star }&=200\left(f\!\left({\frac {Y}{Y_{\mathrm {n} }}}\right)-f\!\left({\frac {Z}{Z_{\mathrm {n} }}}\right)\right)+128\end{aligned}}$

(talk) —Preceding undated comment added 01:12, 21 January 2020 (UTC)

Inconsistency of range

Article reads: "Here, Xn, Yn and Zn are the CIE XYZ tristimulus values of the reference white point ..." stating values 95.047, 100, 108.883.

But in article CIE_1931_color_space the X,Y,Z values seem to be in range 0..1 (or rather that article is missing the information about range of X,Y,Z values) and obviously they cannot be divided by 95..108 to get any meaningful value. This article is missing information, on which scale are the values 95.047, 100, 108.883 - probably on scale 0..100, possibly 0..255 ? 2016-04-18 P.A.S.

The scale of the tristimulus is arbitrarily chosen so that the luminance Y = 100. --Fylwind (talk) 22:48, 8 November 2016 (UTC)

Lab vs. L*a*b*

Could someone revise the lead paragraphs to make clear whether there is any distinction between Lab and L*a*b*? The nomenclature seems to switch between the first and second paragraphs without explanation. What meaning is ascribed to the asterisks? 38.67.192.226 (talk) 14:55, 4 July 2016 (UTC)

Lab color space#CIELAB explains that the asterisks/stars are used to differentiate the CIE L*a*b* from the older Hunter Lab color space. I agree this could be clarified better and the consistency could be improved. --Fylwind (talk) 18:47, 8 November 2016 (UTC)

CIELCh: h = 0?

The text reads "h = 0 means the achromatic colors, that is, the gray axis.", while in the sentence before it defines h = 360° as red. Is this supposed to read "C = 0 means the achromatic colors, that is, the gray axis.", or that any values with C=0 are defined as having h=0°, since they would be the same regardless of the value of h? 141.20.48.10 (talk) 15:07, 7 October 2016 (UTC)

Yes, by definition h=0 and h=360 are not the same in the CIELAB-LCh space. h=0 means the grey axis and h=360 means red. --Farbenprofi (talk) 18:28, 26 February 2017 (UTC)

This must be wrong. I cannot find any reference from a CIE publication that would indicate a noncontinuity at point h=0/h=360; the CIE publication 15.3, section 8.2.1.2 reads:

CIE 1976 a,b (CIELAB) hue angle: h_ab = arctan (b*/a*) (8.13)

and

CIELAB hue angle difference: ∆h_ab = h_ab,1 – h_ab,0 (8.18)

If the line joining the two colours crosses the positive a* axis, Equ. 8.18 will give a value outside the range ±180º, in this case, the value of ∆hab must be corrected by adding or subtracting 360º to bring it within this range

Both of these sections imply a continuity at h=0.

2A01:2A8:8104:C501:228:F8FF:FE25:AA30 (talk) 17:46, 26 March 2017 (UTC)

Yes, thanks for fixing that. Dicklyon (talk) 19:28, 26 March 2017 (UTC)

Revision

Dear all, 2 years ago I made a deep revision of the german CIELAB article. Now I see this is even more necessary in the english one. I would like to try and set focus on following points:

• make better understandable the basics (Spectrum - CIExy - CIELAB)
• take off description of other colour spaces, do not mix
• take off technical sentences where possible
• make uniform the notation to the most common standard (I prefer "CIELAB") and write some words about the many other existing variants (CIEL*a*b*, Lab colors, CIELAB), same for LCh° (LCH, HLC, HCL)
• section usage, where is CIELAB used (software, color measurement, recipe formulation)
• section advantages and disadvantages, improvement ideas
• more well-explaining pictures
• add links to source codes

I will begin and hope my changes will be possible, even if I will reorganise and take off.. - it is important: CIELAB will in my eyes be the most important color definition in future, and problems with CIELAB arise mainly from the fact that it is hard to understand. --Farbenprofi (talk) 19:00, 26 February 2017 (UTC)

Link

Why does "CIEL*a*b* 1976 color space" in the first paragraph link to the CIELUV article? I thought CIELAB and CIELUV were separate? SharkD  Talk  04:55, 10 March 2017 (UTC)

Also, I still can't understand whether this article is supposed to talk about just CIELAB or Hunter Lab or both? SharkD  Talk  04:56, 10 March 2017 (UTC)

In 1976, the CIE simultaneously adopted L*u*v* and L*a*b* because they could not find agreement to select one over the other. So these two color spaces are frequently considered and mostly confused with each other.Jacarter3 (talk) 20:07, 17 October 2018 (UTC)

CIELAB images in article

CIELAB color space top view

CIELAB color space front view

I don't understand what the colored dots in these two images are supposed to represent. Are they the extents of the visible spectrum? I thought CIELAB was infinite in the horizontal directions? This makes no sense. SharkD  Talk  06:05, 10 March 2017 (UTC)

@Farbenprofi: Please explain.

Hi SharkD, The dots are appr. 2000 examples of the boundary of the CIELAB colour space which is not infinite in horizontal directions. (The Lab coordinate space is infinite, but an Lab of 20 1000 1000 makes no sense, it does not represent a color but a nothing). The outest possible colors are those of the biggest possible colour saturation (chromacity) in in the respective lightness area. In real life these are the rainbow colors and combinations of them - those spectrums with a maximum (100%) remission in one or several spectrum intervals.

Following this idea, I took the spectrum 400-700nm in 10nm intervals, set these intervals either to 0% or to 100% remission and calculated the Lab values from these spectrums. You can navigate through and turn around the in Digital Colour Atlas v5 (even demo), window "3D Color Space", selection "CIELab colour space". Holger --Farbenprofi (talk) 16:26, 10 March 2017 (UTC)

Are you saying the dots represent the spectral locus? SharkD  Talk  19:01, 10 March 2017 (UTC)

Any colour is originally a combination of different remissions between 400-700nm wavelenght. You can calculate the Lab value from each wavelenght distribution and show this in the 3dimensional cartesian coordinate systems with the axis L, a, b. The dots are Lab examples of colors with 100% or 0% remission in one or more 10nm wavelenght intervals.

It was a "yes" or "no" question. I don't understand your English, it is confusing. I don't think "remission" means what you think it means. SharkD  Talk  12:32, 12 March 2017 (UTC)

Sorry for my english. The base information of a color is its spectrum, this is a (remission) curve. Remission: how much per cent of the light of a specific wavelength is reflected from the surface. For a spectral curve, you can calculate its Lab values and show this in a 3dimensional coordinate system (L,a,b). The dots show the Lab locusses of the spectral curves of the outest possible colors. Different spectrums can result in the same Lab values (metamerism), therefore a "spectral locus" is not uniquely defined. --Farbenprofi (talk) 13:18, 13 March 2017 (UTC)

Are you trying to create the same plot as the top image here? The scaling seems a bit off, but the shapes seem similar. SharkD  Talk  19:16, 16 March 2017 (UTC)

I am trying to learn how to plot this shape as well, but want to connect the dots with triangles instead of just using dots by themselves. SharkD  Talk  05:44, 17 March 2017 (UTC)

The shapes should be the same. There are similar images, for example http://cielab-farben.de/fragen_und_antworten.html (image above), I do not know how Bruce and others created the triangles leading to the outer surface and shape. I could give you the relevant Lab coordinate table, plz tell your email. Btw. Bruce's terminus "Gamut" is in my eyes not correct for the Lab color space, because a gamut means the possible part of the color space under specific output conditions, but the Lab space does include all possible colors - no specific output condition but from the point of human (standard observer) perception.--Farbenprofi (talk) 09:52, 17 March 2017 (UTC)

Maybe he uses the term "gamut" because there are also many impossible (or invisible) colors in Lab space. I think "Lab space" just means the coordinate system, not the color values plotted in the coordinate system. SharkD  Talk  15:39, 17 March 2017 (UTC)

I would like to look at the table of values, or the code used to generate the values. Thanks. SharkD  Talk  15:45, 17 March 2017 (UTC)

Do you have XYZ versions of these values? It is a little more convenient than LAB. SharkD  Talk  16:14, 17 March 2017 (UTC)

I think that by remission he meant reflection or emission; essentially the spectrum of a source or a reflection of a uniform-spectrum source. There is a theorem somewhere that says the most extreme combinations of lightness and saturation can all be achieved by parameterizing spectra to have just one or two regions of 1s and the rest 0s, like this: 111111111, 0011110000, 1000111111, etc. So you can parameterize these by two values: the wavelength where 0 changes to 1 and the wavelength where 1 changes to 0. Adjacencies in that parameterization will define your surface mesh for plotting. Dicklyon (talk) 19:38, 26 March 2017 (UTC)

A little late, but @Farbenprofi says "Remission: how much percent of light (@ λ) is reflected..." I believe the Eng term is reflectance, which is not quite "reflection". From Wikdipedia: "The dependence of reflectance on the wavelength is called a reflectance spectrum or spectral reflectance curve." "re-emission" wb etymologically the same thing, absent something like fluorescence. Captain Puget (talk) 16:59, 10 June 2022 (UTC)

I like the new graphics from SharkD, but I would egalize the scaling of a/b and L - thus the shape of the color space would be the same as it is in the 2 graphics below. SharkD, can you provide the coordinate values? How did you find them?--Farbenprofi (talk) 20:32, 26 March 2017 (UTC)

No. I am creating a whole series of images using the same basic dimensions. I do not want to change the dimensions for each image. SharkD  Talk  23:04, 26 March 2017 (UTC)

Definitions: coordinate space, color space, gamut

In my eyes, much misunderstanding and not-understanding of CIELAB is caused my the misunderstandable usage of the above words. They should be clearly defined as follows:

Lab coordinate space: all possible combinations of L, a, and b. Many of these are no colors, even if Photoshop shows them as a color. Example: Lab 0 100 100 is shown as a dark red but this makes no sense, a black can not have no color nor chromaticity. Lab 0 0 0 is the only possibility. Photoshop and other programs allow Lab (0...100, -128...127, -128...127) which forms a cylindrical coordinate space. But only appprox 30% of this space are visible colors. The rest is not colors but a mathematical construction which has no counterpart in reality.

CIELAB color space: The part of the Lab coordinate space which is visible colors. See the images in article.

Gamut: The part of the CIELAB color space which can be produced by a certain output method. CMYK FOGRA39, sRGB, AdobeRGB etc.

Clear usage of these definitions would increase comprehensability of CIELAB. --Farbenprofi (talk) 20:55, 26 March 2017 (UTC)

"CIELAB color space: The part of the Lab coordinate space which is visible colors."

This changes depending on the white point you use. It is not one thing. SharkD  Talk  23:01, 26 March 2017 (UTC)

The white point mirrors the light source. D50 and D65 are common. Else, an Lab value (and the whole CIELAB color space) depends on the standard observer: 2° after CIE 1931) or 10° after CIE 1964 Thus, it could be reasonable to create several CIELAB color space graphics? (I think no, the differences are not big.) In each case, the CIELAB color space is only a small part of the Lab coordinate space. --Farbenprofi (talk) 15:17, 27 March 2017 (UTC)

SharkD, you dont mind if your image subtitles will be changed following above mentioned rules?--Farbenprofi (talk) 15:36, 27 March 2017 (UTC)

"Gamut" may be the wrong word for some of these images, but I also don't think we can call each image a "color space", since they can vary depending on the white point used. I think the color space is the coordinate system. Maybe we need more input/feedback from other people. SharkD  Talk  03:07, 28 March 2017 (UTC)

I suppose we could say something like, "CIELAB D65 Color Space". SharkD  Talk  03:18, 28 March 2017 (UTC)

I reread the article, and it says,

"In addition, many of the 'colors' within Lab space fall outside the gamut of human vision, and are therefore purely imaginary; these 'colors' cannot be reproduced in the physical world."

This can be interpreted as meaning CIELAB space includes the colors we can't see. It also calls the limit of human vision a "gamut". SharkD  Talk  03:22, 28 March 2017 (UTC)

Yes, that's pretty confused. An imaginary color is one that does not correspond to any nonegative spectral energy distribution, whether or not mapped through the human 3d subspace of colors. Dicklyon (talk) 04:43, 28 March 2017 (UTC)

The article Impossible_color#Imaginary colors seems to confirm the usage in this article. It says that imaginary colors do not reflect spectral energy distributions or actual physical properties/samples, but rather are just mathematical abstractions used to define a color space like CIELAB. I quote,

"One type of imaginary color (also referred to as non-physical or unrealizable color) is a point in a color space that corresponds to combinations of cone cell responses in one eye, that cannot be produced by the eye in normal circumstances seeing any possible light spectrum.[2] Thus, no object can have an imaginary color. But such imaginary colors are useful as mathematical abstractions for defining color spaces."

Again, "color space" seems to be the mathematical coordinate system, here, and not the physical samples. SharkD  Talk  22:39, 28 March 2017 (UTC)

"Imaginary color" is contradictory, a non-possible thing, like a living dead body or a blueish yellow. Lets better say "Lab values that do not represent a color" or similar. My idea is these differentiations:

• Lab coordinate space, Lab space: all Lab combinations
• Lab color space, CIELAB color space, D65 Lab color space, D50 color space
• sRGB gamut, CMYK gamut, FOGRA39 CMYK gamut,...

With the last, the expression "CMYK gamut" is common, even though there are many different CMYK gamuts are existing. Same with the second. We can say Lab color space, even though there are different ones. (We can also say "VW Beetle" even though there are different ones.)--Farbenprofi 07:59, 28 March 2017 (UTC)

Are there external sources we can use as a reference, or are we just deciding on our own without further research? SharkD  Talk  22:32, 28 March 2017 (UTC)

Regardless, I'm still leaning toward the idea that if there are physical samples, it's a "gamut". If it's just the mathematical coordinate system, it's a "color space". SharkD  Talk  23:08, 28 March 2017 (UTC)

The article Gamut also mentions the "visible gamut" and the "gamut of natural colors", as well. A "gamut" is simply a "complete range" of something. There is no reason it shouldn't be applied to visible colors. SharkD  Talk  20:45, 30 March 2017 (UTC)

A Google search for "visible gamut" produces some results, though not a huge number. How do you search for the alternative point of view? SharkD  Talk  21:24, 30 March 2017 (UTC)

My conviction: "imaginary color" is a non-possible thing, it creates confusion. Color is a imagination in us, but "imaginary color" should mean the contrary (a not possible imagination)? Same with "CIELAB color space" if it should include no-colors ("imaginary colors"?). A color space constists of colors, an Lab 10 100 100 (example) is not a color, but only a mathematical construction without any real meaning, as there is no spectral distribution leading to these values. Summary: we should avoid the term "color" if it can mean a thing outside the CIELAB color space shape. We should name the whole (unlimited) space "CIELAB space" or "CIELAB coordinate space" or "Lab space". IN other words: In an enyclopedia we should not name a man an "imaginary dog"... ;-)--Farbenprofi (talk) 08:53, 30 March 2017 (UTC)

"the four physiological elementary colors"?

This is a completely meaningless statement as far as I can tell. It needs to be reworded or clarified. — Preceding unsigned comment added by 50.207.115.58 (talk) 16:55, 2 May 2017 (UTC)

Lead needs edit

The lead (lede) claims:"The Lab color space describes mathematically all perceivable colors..." This is false for several reasons. It isn't even accurate if we change it to "The Lab color space describes mathematically all colors perceivable by adult humans...". Some people can see into the UV. Some people have more than 3 types of cone cells in their eyes, some people have significant differences in the proteins in the pocket surrounding the retinal (hence different response to light) compared to the "prototypical" model eye, etc. etc. And if it has been established that a newborn is able to "perceive" no more "colors" than adults, I've yet to see it. The Lab (or preferably the L^*a^*b^*) space is based on the CIE color space which was created by a bunch of mostly European, male, white, mature adults working with color - hardly a "random" sample. (They used their own eyes to establish the "averages".) Finally, there are plenty of colors that are "outside" this spectral color space. (According to Wikipedia's article on Color, "As many as half of all women are retinal tetrachromats." - meaning they have 4 retinal pigments hence have 4 "elementary" cone types, rather that the three used in most trichromatic models. Then recall that "color" is a MENTAL perception stimulated by, but not defined by, the reception of light in the eye. (gold, silver, metallics, pinks, white, black, red-green, purple, day-glow/fluorescent,...). Finally while 3 vertices can be drawn on the page and areas of all combinations depicted (i.e. RG, RB, GB, RGB); it is not possible to do that with 4 colors (i.e. 4 elementary colors would require a 3D graph). Anyway, this statement is false.174.130.48.221 (talk) 20:45, 31 August 2017 (UTC)

We argued this above. I think Lab space is infinite in the horizontal directions, just like any Cartesian coordinate system. No, we cannot see an infinite assortment of colors. But then, no two people see the same colors 100% of the time, much less two different species of animal. Further, in practical applications, sometimes calculations need to be made involving colors lying outside the visible locus. SharkD  Talk  00:29, 14 November 2017 (UTC)

Whitepoint & Color Temperature?

I'm a little confused. For reflective surfaces, the conversion from reflectance spectrum to CIE XYZ weights the XYZ functions with the spectrum of the illuminant. That (I think) has the effect of making a perfect white surface (unity reflectance spectrum) have an XYZ coordinate of (0.966797, 1.0, 0.825188) under a D50 illuminant. The mapping from XYZ to L*a*b* requires a whitepoint. If we choose D50 again, we get L*a*b*=(100,0,0). But what if we are converting from sRGB, with a whitepoint of D65? Do we then use the D65 normalization to convert to L*a*b*, in which case #ffffff -> (100, 0, 0)?

If we want to convert a D50 L*a*b* value to sRGB, then it seems like we'd use the D65 normalization, in which case a L*a*b* value of (100,0,0) becomes #ffffff, which I guess approximates what that object would look like under D65 illumination, right? Alternately, if we use D50 normalization to map L*a*b* to XYZ, we get (100,0,0) -> (0.966797, 1.0, 0.825188), which would have the same chromaticity on a D65 sRGB display as a white object under D50 illumination. Is that correct? It seems like a critical but surprisingly subtle distinction. —Ben FrantzDale (talk) 14:13, 13 November 2017 (UTC) (Edited)

I don't recall which color spaces use which illuminant (typically), except for sRGB which is D65. Regardless, if two color spaces use two different illuminants, then yes you are going to have to make adjustments. SharkD  Talk  00:37, 14 November 2017 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified 2 external links on Lab color space. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added archive https://web.archive.org/web/20070701020019/https://partners.adobe.com/asn/developer/pdfs/tn/TIFF6.pdf to http://partners.adobe.com/asn/developer/PDFS/TN/TIFF6.pdf
• Added archive https://web.archive.org/web/20080725151421/http://www.adobe.com/products/creativesuite/pdfs/cscolormgmt.pdf to http://www.adobe.com/products/creativesuite/pdfs/cscolormgmt.pdf

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 5 June 2024).

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 03:17, 15 December 2017 (UTC)