Talk:CIELUV

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Color (Rated Start-class, Mid-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.
 Mid  This article has been rated as Mid-importance on the project's importance scale.
 

Name of this article[edit]

The name of this article is currently "CIE L*u*v* color space." I feel it would be better to base the name on one of the two names sanctioned by the CIE, which I believe to be either "CIE 1976 L*, u*, v* space," or "CIELUV". Lovibond 19:18, 18 May 2007 (UTC)

Is there somewhere on the CIE website or elsewhere that explains this? I'll defer to you on the appropriate article title. --jacobolus (t) 19:50, 18 May 2007 (UTC)
CIE charges money for their publications; it's in Publication 15. The most recent version is the 3rd edition (15.2004). I will refer to either this or 15.2. I know I would prefer the shorter name for the article, if that's okay with you. Good job getting this started, BTW. Lovibond 21:37, 18 May 2007 (UTC)
Okay, in this paper (pdf, “CIE Fundamentals for Color Measurements”), by a guy from the NIST for a conference in 2000, it says “CIE 1976 (L*u*v*) color space” or “CIELUV color space”. I'm happy with “CIELUV color space”, or even “CIELUV” as a title. I think “CIELUV color space” is maybe best then. Then the intro can say something like:
“…the CIE 1976 (L*, u*, v*) color space, also known as the CIELUV color space…”
--jacobolus (t) 08:15, 19 May 2007 (UTC)
The same with all other color spaces. The short forms also used by the CIE are CIELAB, CIELUV, CIEXYZ,... The long forms with year of presentation look like CIE 1931 (Y,x,y), CIE 1964 (Y,u',v'), also like CIE 1931 Yxy and CIE 1964 Yu'v'.
--Albedoshader (talk) 21:00, 12 March 2008 (UTC)

u'v' chromaticity diagram?[edit]

I don't really know too much about the subject, but I just directed a link here from color temperature about the CIE 1976 u'v' chromaticity diagram, which seems like it would fit in reasonably well on this article. Maybe it would also be a good idea to put something about the CIE 1960 uv chromaticity diagram in the history section? --jacobolus (t) 00:45, 15 October 2007 (UTC)

CIE 1960 UCS done and chromaticity diagrams on the way. --Adoniscik (talk) 16:37, 12 March 2008 (UTC)
The correlated color temperature is directly calculated from the CIE 1960 UCS diagram. If you draw a line in this diagram that is both perpendicular to the black body locus and goes through the illuminant color coordinate then the temperature of the intersection point with the bb-locus is the correlated color temperature of the light source. This only works in the 1960 UCS diagram - with the coordinates u and v. So linking to any other diagram than 1960 UCS is likely to confuse readers, especially those who are not very familiar with colorimetry.
--Albedoshader (talk) 21:15, 12 March 2008 (UTC)
Added both diagrams. Yes check.svg Done --Adoniscik(t, c) 20:16, 29 March 2008 (UTC)

Range of L, u and v?[edit]

What are the smallest and what are the biggest values for L, u and v? —Preceding unsigned comment added by 139.18.202.33 (talk) 13:03, 4 December 2007 (UTC)

For what, spectral colors? Consult the chromaticity diagrams in the EL for u and v. For typical gamuts, u* and v* range from ±100 (Poyton, 2003, pg. 227). L ranges from 0 to 100 since Y<Yn .--Adoniscik (talk) 16:37, 12 March 2008 (UTC)

The reverse transformation[edit]

Is there any reason why the inverse formula for X is written as it is? It can be simplified: 
X = 9Yu' / (4 v') Hpablo (talk) 22:26, 4 February 2008 (UTC)

Yes check.svg Done --Adoniscik (talk) 16:37, 12 March 2008 (UTC)
As far as I know the Lightness is calculated the same way as for CIELAB:
L^{*} = 116 \left(\frac{Y}{Y_n}\right)^{\frac{1}{3}}-16 , if \frac{Y}{Y_n} > 0{,}008856
L^{*} = 903{,}3 \left(\frac{Y}{Y_n}\right)^{\frac{1}{3}} , if \frac{Y}{Y_n} \leq 0{,}008856
or better:
L^{*} = 116 \left(\frac{Y}{Y_n}\right)^{\frac{1}{3}}-16 , if \frac{Y}{Y_n} > \frac{216}{24389}
L^{*} = \frac{24389}{27}
 \left(\frac{Y}{Y_n}\right)^{\frac{1}{3}} , if \frac{Y}{Y_n} \leq \frac{216}{24389}
CIELUV is based on CIEXYZ and CIEYU'V' (for white point), not on CIEUVW. Even Bruce Lindbloom uses the above mentioned formulae.
These would also make more sense, because LUV and LAB were created in the same year. Why should the CIE measure lightness in two different ways at the same time? If the goal is perceptual uniformity then it's the logical consequence that lightness is weighted the same in both spaces.
--Albedoshader (talk) 20:16, 12 March 2008 (UTC)
At some level everything is based on CIEXYZ. CIELUV is closest to CIEUVW; both use similar definitions of chromaticity and lightness. The u'v' chromaticity diagram (CIE 1976 UCS) certainly is similar to MacAdam's uv diagram (CIE 1960 UCS). I had not heard of "CIEYU'V'"; aren't u' and v' defined in CIELUV? (Schanda says so on Colorimetry, page 64.) The only references to it I could find were derived from the German WP. I posted an explanation about the lightness formula (it is correct, but good catch). Adoniscik (talk) 21:20, 12 March 2008 (UTC)
I don't doubt Schanda. But why is the only form I could find in papers the one above? For CIE YU'V' see for example this PDF from Alan Ford and Adrian Ford about color space conversions:
http://www.poynton.com/PDFs/coloureq.pdf
--Albedoshader (talk) 22:04, 12 March 2008 (UTC)
I guess it depends on which version of publication 15.2 they read (I don't have it myself). As the quote from Schanda says, there are several versions which say the same thing. I think the fractional version is more elegant. Regarding YU'V': its meaning is obvious enough, but I have not seen it mentioned formally (books and papers). --Adoniscik (talk) 22:13, 12 March 2008 (UTC)
Lindbloom proposed this solution to get rid of the discontinuity and Schanda answered him the CIE was adopting it:
Note: I was notified by Dr. János Schanda and Todd Newman on 3 April 2003 that the CIE 15.316 and the CIE Standard on CIELAB will implement this fix (i.e. using rational rather than decimal values for these constants). (http://brucelindbloom.com/LContinuity.html)
Regarding YU'V': I think you're right. Most probably it's simply an interim color space developed on the way to CIELUV. Like the CIECAM97s and CIECAM97c models as interim solutions between CIE94(and other works) and CIECAM02. But at least it is the better choice as chromaticity diagram, compared to CIE1931 and CIE 1960 UCS.
Regarding my first comment: Forget it, the form in the article is mathematically the same, just in a slightly different form. I should have seen it earlier.
--Albedoshader (talk) 22:43, 12 March 2008 (UTC)

Color space images[edit]

I'm very active in the German colorimetry section and I'm working on 3D-animations of the optimal color bodies for all color spaces. It's no problem to put them here too, with English captions. The animations show rotations around the color spaces. The spaces themselves are cut into slices in regular lightness steps from 5% to 95% to optimally show the change of shape with lightness. Here's an old example I made during the development of the slicing algorithm: http://www.albedo-cg.de/CIE/Yxy_lightness.gif For up to date animations check my German user dicussion page every few days: http://de.wikipedia.org/wiki/Benutzer_Diskussion:Al%27be:do. The calculation process itself is quite time consuming, so the completion will take some time. But the results are very rewarding. Spaces I'working on:

  • CIEYxy (2° observer)
  • CIEYu'v' (10° observer)
  • CIEUVW
  • CIELAB
  • CIELUV
  • DIN99

--Albedoshader (talk) 11:42, 12 March 2008 (UTC)

Very good. I already used one of them to illustrate optimal surfaces in gamut. Can you prepare versions without any text so the other WPs can use them too? --Adoniscik (talk) 16:37, 12 March 2008 (UTC)
Yep. Stuffing the info into the image name should work. By the way, the picture you mean shows a spectrum of an optimal color, not of an optimal surface. I've already changed the caption.
--Albedoshader (talk) 19:46, 12 March 2008 (UTC)
Isn't it a property of the surface (as opposed to the light source)? Adoniscik (talk) 20:01, 12 March 2008 (UTC)
Optimal colors already incorporate both the surface reflectance/transmittance spectra and the light source spectrum. The resulting spectra that lead to the color are the dot products of both the reflectance/transmittance and the light source spectra. Each color contributes only one point to the color body surface. To build a complete optimal color body you need an infinite amount of points, and thus an infinite amount of reflectance/transmittance spectra or colors.
I think you misunderstand what is meant with the concept of an optimal color body/surface. Optimal colors are purely theoretical. No physical material can reflect or transmit an optimal color spectrum. All real world (reflective/transmissive) colors are within the boundaries of the optimal color body which is not a physical surface, but a purely mathematical entity. The optimal color body depicts the absolute limits of realizable (reflectance/transmittance) colors in an arbitrary color space - nothing else.
There are color bodies and body colors(I hope this is the proper English term, I did simply translate it from German). Every color that results from an interaction of light with a physical material (reflection, diffusion, transmittance...) is a body color. The optimal color body (mathematical) is the sum of all body colors (physical). The other type of color is called light colors, a rather self-explanating term. Light colors are all colors realizable by light sources.
An optimal color body looks like this (cut into slices from 5 to 95, stepwidth 10):
Optimalfarbkörper CIE1931 y5-y95-illuminant f4-100dpi.gif
I very much hope that the terms I've introduced are the correct English ones.
--Albedoshader (talk) 20:51, 12 March 2008 (UTC)
Thanks a lot for taking the time to correct my understanding. The proper term might be color solid; at least, that is what Martínez-Verdu et al (2007) uses. (de:Körperfarbe?) --Adoniscik (talk) 21:48, 12 March 2008 (UTC)
Yeah, solid was the word, my memory isn't the best at the moment ;) But color solid is Farbkörper (lit.:colorbody/ mathematically: colorsolid) in German. Ah, I found the term for Körperfarbe(n)! Simple as hell: "surface color(s)".
Thanks for your help!
--Albedoshader (talk) 22:28, 12 March 2008 (UTC)

Adams chromatic valence[edit]

I think the article lacks a strong relation to the Adams chromatic valence concept, as it's claimed in the preface section. For example, look at the Hunter Lab which is clearly expressed in terms of Adams theory. Not only we see that Hunter Lab is UCS-based and luma-scaled, but that it also implies opponent process, which is not the case with CIELUV.

In fact, the referenced book “Color appearance models” by Fairchild in section 10.5 explicitly blames CIELUV for using subtractive shift to reference white rather than pseudo von Kries chromatic adaptation by XYZ-normalization, which is involved in CIELAB. May be I am missing something here, but no simple explanation pops up for CIELUV becoming an opponent process model. Of course, u'v' scales somehow resemble “green-red” and “blue-yellow”, but isn't it just a coincidence? I've never met any mentioning that UCS was designed with opponent process in mind, even if it was intended just to mimic it rather that to fully implement.

213.234.235.82 (talk) 15:09, 19 June 2008 (UTC)

It is a coincidence. CIELUV is a simple transformation of the CIE1931 Color space.
I've made several 3d-renderings of the overall shapes. See for yourself:
CIE1931 (Yxy) color space:
de:image:Optimalfarbkörper_CIE1931_y5-y95-illuminant_f4-100dpi.gif
CIELUV color space:
de:image:Optimalfarbkörper_LUV_5_95-D65-100dpi.gif
You can clearly see the similarities.
--Albedoshader (talk) 18:26, 28 June 2008 (UTC)
CIELUV is very much an Adams Chromatic valence space, because it is constructed from a uniform lightness scale and a uniform chromaticity diagram, and the chromatic components are computed by multiplying chromaticity by lightness. That is the essence of an Adams chromatic valence space. I do not understand your (user 213.234.235.82) mention of opponent processes. These are not an explicit component of the valence space. Nor do I believe the use of a Helson-Judd style white point adjustment affects CIELUV's status as a valence space. The essential aspects are, as I mentioned earlier, a uniform lightness scale and two chromatic coordinates computed by multiplying lightness by uniform chromaticity. CIELUV has both. Lovibond (talk) 17:23, 28 April 2009 (UTC)

Correlate/Correlation[edit]

I've reverted an edit that had identified C* and h as chroma and hue correlations; the previous version identified them as correlates. The latter is consistent with customary and accepted practice. Lovibond (talk) 17:29, 28 April 2009 (UTC)

"Chroma" link[edit]

The link to chroma under Cylindrical representation currently takes the reader to a disambiguation page. I don't know enough about the subject to make the link more specific, but I'd think that one of colorfulness, chrominance, or Munsell_color_system#Chroma would be the right target. Scott Pakin (talk) 18:06, 25 July 2010 (UTC)

colorfulness is the right link, since that page describes (at least nominally) colorfulness, chroma, and saturation. –jacobolus (t) 18:55, 25 July 2010 (UTC)

Y_n?[edit]

The transformations reference Y_n but it is never explained what this variable is. Can someone remedy that? Is it the Y tristimulus component of an XYZ white point (e.g. D65's Y=100.00?). Thanks! 174.252.49.198 (talk) 20:26, 3 November 2010 (UTC)

Yes, Yn is the value of Y for the white point. The idea is that Y/Yn is in the range [0, 1]. –jacobolus (t) 21:00, 3 November 2010 (UTC)

Asterisk or not[edit]

In the line "The cylindrical version of CIELUV is known as CIE LChuv, where C*uv is the chroma and huv is the hue:[6]", doesn't the h miss a "*"? I don't have access to the original citation. — Preceding unsigned comment added by OxygenBlue (talkcontribs) 20:41, 19 May 2013 (UTC)