Talk:Chromaticity

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Color (Rated Start-class, High-importance)
WikiProject icon This article is supported by WikiProject Color, a project that provides a central approach to color-related subjects on Wikipedia. Help us improve articles to good and 1.0 standards; visit the wikiproject page for more details.
Start-Class article Start  This article has been rated as Start-Class on the project's quality scale.
 High  This article has been rated as High-importance on the project's importance scale.
 

This article has comments here.

merge with saturation?[edit]

This should really be merged into Saturation: Colour theory, this would end a stub, and join it to its "father" article. It would be better if it was a "sub-chapter" to Saturation: Colour theory.

I'm not sure I agree. I'd like to include a bit about the meaning of chromaticity in accelerator physics (specifically in storage rings), and that wouldn't fit into Saturation: Colour theory at all. Unless a separate page was created... SteveM 09:06, 24 July 2006 (UTC)

In regard to color theory this page is erroneous. The term Chromaticity encompasses both saturation and hue as a single concept, not purity as the article describes. An illuminant or color has both Chromaticity and Luminance. The chromaticity component (x,y) describes a normalized tristimulus ratio (X,Y,Z.)

I do not think this page should be moved. I think a stub should be placed here and I would be willing to work on the color portion of this page. Various Chormaticity diagrams could all be located and explained on this page.

CIE 1931 x,y - CIE 1976 ū,v - Should both be described and depicted on this page.

Illuminant 17:28, 17 October 2006 (UTC)

Strongly oppose. Chromaticity encapsulates saturation and hue. It should stay here so we can discuss chromaticity diagrams, and perhaps perceptual uniformity. --Adoniscik (talk) 16:06, 4 February 2008 (UTC)
This 2006 is proposal is no longer active (no tags are on the articles). Dicklyon (talk) 16:11, 4 February 2008 (UTC)

Chromaticity diagram colors are wrong[edit]

The CIE 1931 x,y chromaticity space diagram at the top of the page is incorrectly colored.

Note that white is far below the blackbody curve. Now, it's possible the author bizarrely chose to use a white-point of say Illuminant E, rather than web standard sRGB's D65, or even print's D50. But almost always, the gap is just an easy indicator of an incorrect white-point calculation.

This is not uncommon. Consider the diversity in google image search for "cie". Many good sites have (variously) wrong diagrams. Problem is, several places on the web provide incorrect software, incorrect algorithm descriptions, or both. So people keep getting burned.

Other articles containing diagrams with incorrect chromaticity include Color temperature, Black body, and Planckian locus.

Fyi. 98.216.110.149 (talk) 00:25, 4 September 2009 (UTC)

Hi - There are a number of chromaticity diagrams that I generated, all derived from the original one at http://en.wikipedia.org/wiki/File:CIExy1931.png. This particular one has an explanation of how the diagram was generated and it reads:

The colors for this diagram were generated using the RGB color space in en:Adobe photoshop. The transformation from xy chromaticity coordinates was done using the en:sRGB color space specification on the [X,Y,Z]=[x,y,1-x-y] tristimulus values, then multiplying by a constant so that one of the R, G, or B values was maximized. Assuming that one's monitor converts Adobe photoshop RGB according to the sRGB color space (probably a good assumption) then, within the sRGB gamut, the chromaticities are correct, but are incorrect outside the gamut. (See the sRGB article for a description of the sRGB gamut). The process of maximizing the value of R, G, or B results in a distinct 3-pointed star in the diagram, centered at the D65 white point. This is because, although the chromaticities are correct, the luminosities (brightness values) are not equal across the diagram. If the luminosities were all made equal, then the entire diagram would be rather dark, since pure blue has a low luminosity. Any attempt to equalize luminosity to remove the star will reduce the overall luminosity of the diagram, and the star will not completely disappear until the diagram is very dark. Alternatively, we could blur the colors to get rid of the star, which would give incorrect chromaticities. I have opted for correct chromaticities at maximum brightness, thus the presence of the star.

Now, this may not be the absolute best way to do things. For example, someone has made the very good suggestion that rather than have the out-of-gamut colors be the same as the nearest in-gamut-color that lies on a line between the point and the D65 white point, one should have it equal to the color of the nearest in-gamut point based on a distance measured in a good chromaticity metric, like the Lab color space. At any rate, PLEASE, if anyone decides to change things, give a very detailed description of how the diagram was generated, and try to make it have some connection to reality. PAR (talk) 11:49, 9 September 2009 (UTC)
Also - I just looked at the Planckian locus diagram (April 2005) and it is not using the latest color rendition (June, 2005). I can fix any diagram that has a wrong color scheme, but people keep screwing around with them, trying to make them look pretty, so I would like to do it finally as right as possible. Changes that are actual improvements would be the suggestion above, making it an SVG rather than PNG... Any other suggestions? PAR (talk) 12:02, 9 September 2009 (UTC)

Chromaticity in color spaces[edit]

In the section "In color science," the article currently states, "the chromaticity coordinates are a* and b* in CIELAB, u and v in CIELUV...." Firstly, the CIE has not, to my knowledge, sanctioned or promulgated chromaticity coordinates for the CIELAB color space (if it did, chances are they would be something like a*/L* and b*/L*, which is essentially how it works in CIECAM02). a* and b* do not fit the definition of chromaticity coordinates. Secondly, the chromaticity coordinates in CIELUV are u', v', not u, v, as the article states. Accordingly, I am removing the reference to the CIELAB color space, and correcting the reference to the second. Lovibond (talk) 01:28, 19 January 2013 (UTC)

On second thought, more corrections are necessary, as CIELUV needs to be moved to the second paragraph. Lovibond (talk) 01:31, 19 January 2013 (UTC)
True, a* and b* are not exactly chromaticity; a*/L* and b*/L* would not be either; you can't get there from here. On the other hand, a* and b* function sort of like chromaticity, and I wouldn't be surprised if some sources refer to them as such. If we mention them, it should be in the context of a source and an explanation. I haven't looked to see what it said, but I probably have objection to taking them out. Dicklyon (talk) 01:33, 19 January 2013 (UTC)
"a* and b* are not exactly chromaticity" is an understatement! They're not exactly chromaticity in the same sense that C* is not exactly excitation purity. As far as a*/L* and b*/L*, they do behave much more like chromaticity than a* and b*. Saturation is computed as a distance in chromaticity space, and is defined as colorfulness relative to brightness. And colorfulness is to chroma (which may be resolved into two components, a* and b*) as brightness is to lightness. If a source refers to a* and b* as chromaticity coordinates, it is contrary to the definitions in the International Lighting Vocabulary, CIE Publication 15, Fairchild, Hunt, and other authorities, and therefore of questionable reliability on at least that point. Lovibond (talk) 02:13, 22 January 2013 (UTC)
Good call. I don’t like the section header “in color science” either. It would be nice to first explain that “chromaticity” technically describes a 2-dimensional quantity which is 1–1 related to relative cone responses (or approximations thereof, or linear combinations of those, such as xy, u'v', etc.), under an assumption of linearity (Grassmann’s Law). Then it would be nice to show some chromaticity diagram where relative cone responses are the axes, and a few other chromaticity diagrams, explaining how they arose, explain the advantages (mainly linearity of light combinations?) and down-sides (perceptual non-uniformity [macadam diagrams], hue shift along straight lines, inconsistency between hue purity and perceived chroma/colorfulness, etc.) to using those as color models. It might be a good idea to show how instead of x, y we can write a chromaticity in terms of "dominant wavelength" and "hue purity". It might be nice to show some other diagrams that can be easily drawn on a chromaticity diagram, such as the gamut of diffuse-reflective non-fluorescing surface colors under some specific light source. Then finally it would be nice to contrast with other models which separate luminance/lightness from color in different ways, ideally with some diagrams showing 3-dimensional gamuts, starting with CIELUV, but also maybe CIELAB, Munsell, and e.g. CIECAM02. Then again, writing all that, sourcing it, &c. &c. would be a big bunch of work, and I dunno if anyone has time to do it. :-) –jacobolus (t) 02:20, 19 January 2013 (UTC)
I'm not sure what you mean by "1–1 related to relative cone responses (or approximations thereof, or linear combinations of those, such as xy, u'v', etc.)"; that's pretty squishy. It's not 1–1 with the tristimulus, obviously, but it's 1–1 with some kind of 2D projection; not clear that "relative cone responses (or ...)" captures the essense of that. Depending on how one defines it, a* b* might or might not be a chromaticity. What do you call these mappings that are linear in numerator and denominator? Projections? Projections through the origin? Dicklyon (talk) 03:41, 19 January 2013 (UTC)
The standard term for this construct in mathematics is projectivization. Coordinates with include denominator are called "affine coordinates" or "non-homogeneous coordinates" on a projective plane. Incnis Mrsi (talk) 06:42, 19 January 2013 (UTC)
I shouldn’t have said “relative cone responses”; I’m not sure what the clearest phrasing is. What I meant though was that chromaticity describes cone responses for a particular luminosity. Saying that it’s some linear transformation of a projection of cone responses onto a plane would maybe be better. I don’t think a* or b* (or u*, v*) is ever really “chromaticity”. –jacobolus (t) 04:32, 22 January 2013 (UTC)