Talk:HSL and HSV/Archive 3

Archive 1

Archive 2

Archive 3

Archive 4

Archive 5

Archive 6

Examples of RGB and HSV or HSL images in the main article

In the sample images on the right side of the article -- the Lightness (and Value) image is shown in "redish" coloration. The three HSL components of the RGB image should be in gray level (monochrome images) since each pixel in these images has only a single value. For instance, according to the article, L =1/2(min+max). There are no 3 color components to show a given pixel in color.

BTW, PainShopPro separates color images into HSI (or RGB or CYMB) and these are displayed as monochrome.

Just a suggestion to reconsider the examples.

N. Gat 67.111.26.178 22:32, 10 October 2007 (UTC)

I didn't make any of those channel separation examples, but the several people who did make them (a year ago) made HSV, HSL, Lab, CMYK, RGB, YCrCb, YUV, YDbDr, YIQ, and maybe a few more besides, all of which are colored similarly. --jacobolus (t) 01:36, 11 October 2007 (UTC)

It's possible that the saturation, lightness, and value images are shown in red because in HSV and HLS, red corresponds to a hue of 0 (this was arbitrarily chosen for these systems, as far as I know). Thus, the hue image shows actual hues, and the other images show saturation, lightness, and value with a hue equal to 0. I don't know if that's what the creator(s) of those images had in mind, but it seems likely, especially if they were showing HLS and HSV in context with other color models, where a hue of 0 may be assigned to some other color. Of course, for the purposes of illustration, you might just as well choose to display gray images. It would just be a matter of preference or purpose. Seaandsky 15:54, 11 October 2007 (UTC)

And actually, in this specific case, I'd personally prefer to have hue and saturation shown with colors, and value/lightness/brightness/whatever shown in gray. But the current ones are fine enough that I'm not going to bother changing it myself. :) --jacobolus (t) 16:22, 11 October 2007 (UTC)

I had never really looked at those before, but now I think I see their point. The H is shown as an image of variable hue with a chose (max) saturation and value. The S is shown as a image of variable saturation, for some chose hue and value. And the value is shown as variable value with fixed hue and saturation. The latter could choose saturation 0 and make it gray. But the S image can't be shown without a color, and picked hue 0 for that. Confusing, but with a certain logic. Dicklyon 23:51, 11 October 2007 (UTC)

I think they should be in grayscale, as well. All image creation software does so in this way, as far as I know. SharkD (talk) 20:04, 12 January 2008 (UTC)

The inconsistency of these images bothered me too. That's why I made ones of my own. see my gallery. I'm not confident enough to put them into this article. Maybe one of you knows their way around a page edit :P --Crackwitz (talk) 01:48, 20 August 2009 (UTC)

I wonder whether anyone thinks these (i.e. the original ones, not Crackwitz’s, though the question applies to his just as much I suppose) images were actually especially useful for people trying to figure out how these spaces worked. I haven’t put them back into the re-written article, but instead there are a set of images showing different ways of computing lightness/value/etc., and then some images showing hue rotation. I always felt that trying to demonstrate breaking up an image by hue and chroma/saturation/whatever via a demonstration that normalized all the colors in the image to the same lightness was a rather futile exercise, and more misleading than enlightening. Can anyone think of other decent ways to deconstruct the colors in an image to show what’s going on? It would perhaps be useful if we could show a couple of images along with where their colors fell in the 3D geometry of the color space, or even in 2D slices. I’m not sure exactly what form such a picture would take though. –jacobolus (t) 04:25, 14 February 2010 (UTC)

Barn Image is Confusing

The barn image is confusing. What does it mean it has its RGB or HSL values displayed? More explanation about what is going on in each image would be really helpful. The one on the left is supposed to be R, G, and B, right? Why does one of them have yellow and blue in it? Shouldn't each one only have one color? Maybe break down each image with captions for each one explaining what's going on and what that means. Thanks! Saffolicious (talk) 05:44, 2 January 2010 (UTC)

It has also been suggested that the images be presented in grayscale, as they are in most image editing programs. SharkD  Talk  04:12, 4 January 2010 (UTC)

"Intuitive" HSV/HSL to RGB

Fully saturated colors. SharkD still doesn't "get" it.

I find the HSL/HSV to RGB conversions horribly unintuitive, mostly because the case distinction over H is kept until the end. I would find it more comfortable to have a fully saturated color, then "apply" SV and SL to it, respectively. Both the SV and SL "transformations" do not depend on the hue and are equal for all three RGB channels (I will just use C as placeholder below).

for HSV, that would be:

1. Compute the fully saturated color according to the image on the right.
2. Apply saturation: We basically scale the RGB values toward 1 (white), i.e. the difference to 1 is scaled down, 1 is a fixed point:
   • $C\gets 1-S\cdot (1-C)$
3. Apply value: We simply scale the RGB values toward 0:
   • $C\gets V\cdot C$

and for HSL:

1. Compute the fully saturated color according to the image on the right.
2. Apply saturation: We basically scale the RGB values toward 0.5 (gray), i.e. 0.5 is a fixed point:
   • $C\gets {\tfrac {1}{2}}+S\cdot \left(C-{\tfrac {1}{2}}\right)$
3. Apply lightness. This is the nasty part:
   • for $L\leq {\tfrac {1}{2}}$: The color gets darker, black for $L=0$ and unchanged for $L=0.5$. This means scaling toward 0.
     • $C\gets 2L\cdot C$
   • for $L>{\tfrac {1}{2}}$: The color gets lighter, white for $L=1$ and unchanged for $L=0.5$. This means scaling toward 1. Because we want $L=1$ to yield $C=1$ (i.e. scale 0), the scaling factor is $2-2L$ (and not $2L$).
     • $C\gets 1-(2-2L)\cdot (1-C)$
   • note that for $L=0.5$, both cases do nothing.

Would be cool if somebody could verify these algorithms and eventually include them in the article. Thanks! MoA)gnome (talk) 20:25, 25 July 2008 (UTC)

I'm not sure I understand the chart at right. Why don't the saturation values go from 0 to 1? They seem to instead go from 1/4 to 3/4, if I read the y-axis correctly. SharkD (talk) 02:54, 8 September 2008 (UTC)

The caption appears to be misleading. The range of values has been adjusted based on S and V values less than 1. The illustration is for about V=0.75 and S = 0.7, looks like. Dicklyon (talk) 04:26, 8 September 2008 (UTC)

Yeah, sorry, I took the image from the article without paying attention to the fact that the values are in a smaller interval. What I mean by "fully saturated" would be the curves stretched vertically so they reach from 0 to 1; i.e. the possible "initialization" colors are the ones in the color band above the curves. —MoA)gnome (talk) 22:31, 11 September 2008 (UTC)

On a side note, what I described exactly corresponds to moving through the according color "cylinder" (image also taken from the article). HSV (right) starts at the upper edge, HSL on the outside at half the height. Applying S means moving toward the center, applying V/L, respectively, means moving down or up (there's no moving up for HSV, of course). —MoA)gnome (talk) 22:51, 11 September 2008 (UTC)

If the image accurately reflects what you describe, then what you describe is already covered by the article. SharkD (talk) 23:32, 11 September 2008 (UTC)

I am not saying that this is anything new. My point is just that the HSL/HSV to RGB conversions are quite obscure IMHO; it is hard to tell what saturation and lightness/value do to the quasi "initial color" that I called "fully saturated". The cylinder image does reflect it, though. —MoA)gnome (talk) 21:06, 13 September 2008 (UTC)

Ah, I see. I automatically assumed that the hue gradient at the top of the image was fully saturated/valued. I expect that other readers might also become confused by this. SharkD (talk) 01:23, 29 December 2008 (UTC)

OK, I lied. I still don't get it. SharkD  Talk  04:26, 2 February 2010 (UTC)

---

I’ve tried to put in more intuitive conversions. What do you think? –jacobolus (t) 22:12, 5 February 2010 (UTC)

L*C_a*b*h_a*b*

Maybe L*C_a*b*h_a*b* [edit: per the previous discussion] should be mentioned in the "Comparison with other color models" section? It's fairly similar. I can't find any references to it using Google, though. SharkD (talk) 05:56, 29 June 2009 (UTC)

L*C*_abh*_ab and L*C*_uvh*_uv are just alternate (cylindrical rather than rectangular) representations of CIELAB and CIELUV. They could certainly be discussed, if you like. There are all kinds of references: search for "CIELAB LCh". —jacobolus (t) 05:18, 1 July 2009 (UTC)

I've tried to put a brief description in about these. It still needs some filling out (one sentence currently just trails off!), and could use some pictures. –jacobolus (t) 10:32, 12 March 2010 (UTC)

Device independent image formats

Just a quick question: are there any web-safe device-independent image formats, or is that still several years down the road? SharkD (talk) 16:12, 30 June 2009 (UTC)

I don’t understand the question. What is “web safe”? —jacobolus (t) 05:19, 1 July 2009 (UTC)

Mainly I mean whether or not they are supported by browsers. SharkD (talk) 20:56, 1 July 2009 (UTC)

Why would you think it might be "down the road"? Is there evidence that there's some interest in combining the "web safe" concept with the "device independent" concept? Dicklyon (talk) 00:16, 2 July 2009 (UTC)

Safari and Firefox will accept ICC profiles in images, so that if you put a JPEG or PNG image with an embedded profile on a web page, the color will look (roughly) correct in Firefox or Safari on a properly characterized display. If you put an untagged image up, Firefox will assume that it is sRGB, but Safari will (contra the spec, and contra "proper" behavior, but apparently in the interest of speed and agreement with colors in Flash content) just send the RGB coordinates direct to the display color space. I think Internet Explorer is probably just beyond hope w/r/t such considerations. No idea about Google Chrome. –jacobolus (t) 10:37, 12 March 2010 (UTC)

working on a nearly complete re-write

Hi everyone. At long last I've gotten sick enough of the state of this page that I'm working on a near complete rewrite. While I write it, I have it at Talk:HSL and HSV/replacement in progress, and I'll move it to HSL and HSV when it's ready to go. Feedback about my current outline or already written prose is welcome, but for now, don’t edit that page directly, until I’ve had a chance to write the complete article. I still need to make several more diagrams, among other things. –jacobolus (t) 04:16, 26 January 2010 (UTC)

• Firstly, the article probably should have been created in User space. But it's not a big deal.
• "Both of these representations are in very wide use in computer graphics, and one or the other of them is often more convenient than RGB, but both are also commonly criticized for not actually separating color-making attributes, and for their lack of perceptual uniformity." I understand the perceptual uniformity bit, but not the bolded part. Do you mean the photometric color-making attributes? If so, then you should say so. What do you mean when you say "separating"?
• It should be pointed out that the problems with regard to lack of perceptual uniformity in each of the models is inherited from RGB. I.e. HSL and HSV are perceptually inaccurate because RGB is.
• I like how you created File:Tint-tone-shade.svg. Some analog depictions of some of the other painters' terminology (chromaticity, nuance, etc.) would be nice to have as well.
• The section on "geometric derivation" is a bit convoluted. It needs copy-editing, and needs to be more concise.
• The article looks a bit like it's suffering from image soup. Putting some effort into arranging and aligning the images would help make the article appear a bit more organized.
• Lastly, I think the existing article is better suited for readers with less understanding (i.e. beginners). For instance, readers might want to go directly to the meat of the matter; what the model looks like, how it is used in the professional world, and the conversion formulas. Jumping directly into the derivation material is a bit like telling the recipe for Chicken Kiev before saying what it actually is. SharkD  Talk  00:49, 31 January 2010 (UTC)

Yes, it definitely needs a copyedit. The problems w/r/t perceptual uniformity, etc. still need to be described in detail, in the section "disadvantages". I don't know whether the geometrical derivation section can be made *too* much more concise and still explain how these things work. Great effort *has* been put into arranging the images, so that they flow properly; there might be some way of improving it a bit though. I think that the existing article is terrible for beginners, who are likely to be misled and left confused, since it is vague and imprecise, and does not extremely clearly state the problems with HSL/HSV. You might be right though that a section at the top with the “basic idea” might be helpful. I still have a way to go, in particular with adding images showing various cross-sections and 3-d renders of these models. I don't think the ones in the current article (e.g. near the top) do a good enough job. –jacobolus (t) 01:54, 1 February 2010 (UTC)

I meant the images in the derivation section, sorry. The rest of the article is fine. I think changing the width of the second image to match the first and third would suffice. Also, you might want to temporarily lower your graphics card resolution. I browse at 1024x768 and things are cramped. SharkD  Talk  04:12, 2 February 2010 (UTC)

Yeah, it's a bit of a tricky trade-off. It currently looks best for browser windows of between 900-950 pixels wide. People using 800x600 resolution are definitely going to suffer a bit from the layout, as are people who size their browser windows to wider than about 1100 pixels (their wikipedia experience is just going to suck generally though, because every line of text will be way too wide). –jacobolus (t) 03:52, 3 February 2010 (UTC)

Okay, this is getting more complete. Continuing feedback is welcome (again, the page). –jacobolus (t) 00:27, 5 February 2010 (UTC)

Does anyone have a good idea how I can make this section flow with the images it has, at all monitor widths? This is the best I could come up with, which works for widths 900-1050 pixels, but then leaves a bigger gap the wider the window gets after that. Maybe it's no big deal – people with wider browser windows are going to get a somewhat unpleasant Wikipedia experience regardless. I wonder if anyone knows, e.g., whether I can make images inline elements w/ wiki markup, so that as many will try to fit in each row as possible, flowing around the image above to the right. –jacobolus (t) 16:16, 5 February 2010 (UTC)

As for the images spreading into neighboring sections, I would suggest combining the first two images in the 'Hue and chroma' sections into a single image, like this. Or, split them into three images and use the multiple image template. SharkD  Talk  03:35, 6 February 2010 (UTC)

The images in the 'Motivation' section could be combined into a single group like this. SharkD  Talk  03:53, 6 February 2010 (UTC)

Painters long mixed colors by combining relatively bright pigments with black and white. Mixtures with white are called tints, mixtures with black are called shades, and mixtures with both are called tones. See Tints and shades.

This 1916 color model by German chemist Wilhelm Ostwald exemplifies the “mixtures with white and black” approach, organizing 24 “pure” colors into a hue circle, and colors of each hue into a triangle with corners of the “pure” color, white, and black. The model thus takes the shape of a bicone.

This diagram, from a Tektronix patent filed in 1983, shows the bicone geometry underlying the HSL model.

Colors are combined by adding values. (Should ideally use the same color as in the bottom image.)

The RGB gamut can be arranged in a cube.

But one of the images is intended to flow with the text, while the others are supplementary/incidental. I don't like stacking them the way you have here: it seems extremely forced, and makes things harder to see. There's no reason to not just let images be big enough to read/etc. –jacobolus (t) 07:16, 6 February 2010 (UTC)

A few more points: If you think it’s currently cramped for narrow browser windows, making a 460ish-pixel-wide and hundreds-of-pixels-tall block of images would be much worse. Especially since, in the narrow column next to them, not much text would fit, and so big blocks of text below would necessarily be full-width (and therefore less readable; we end up with the worst of both worlds, with a super narrow column for one part and a super wide one for another). The tint/tone/shade image is as small as it can reasonably be and still be readable, and if it's unreadable it's essentially useless and might as well be removed. I don't really care too much for the RGB cube image. I just put it in because I think we need some decent illustration of a cube, and this one was handy, already in (I think the German?) wikipedia; we could certainly make a different one. The tektronix patent image belongs as near as possible to the text about tektronix; it's a logically separate part of the text from the ostwald/tint-tone images, and sticking them together would be confusing. –jacobolus (t) 07:44, 6 February 2010 (UTC)

Break

Okay, I've started transferring things. –jacobolus (t) 17:21, 5 February 2010 (UTC)

A couple of remarks

• I think the term "color model" should be used instead of "color space" except where specifically required. For instance, in the top image.
• I think that the 2D images of the cylindrical cross-sections (File:Hsl-hsv chroma-lightness slices.svg and File:Hsl-hsv saturation-lightness slices.svg) should be accompanied by 3D images of the actual solids. Not sure how you would create these in SVG. I don't personally own software tools that can do this. There's currently only one image in the article actully showing a 3D representation of the cylinders.
• Please use the same color in all the charts. For instance, in the first set, the color has a hue value of 200°; in the second set the color has a hue value of 230° (and its complement has a value of 50°).

SharkD  Talk  23:32, 5 February 2010 (UTC)

What charts are we talking about? The hue/chroma and saturation sections use the same colors. As far as I know nowhere is a 200° hue used. I don’t think that the “cones” are actually useful to show at all, but I have a friend who is going to render 3d views of both HSL and HSV cylinders. –jacobolus (t) 00:19, 6 February 2010 (UTC)

Oh, now I see what you’re saying. No, the difference is motivated by which best show off the effects we want to show: We pick a point 1/3 the way along an edge so that it’ll show a reasonable difference between H and H₂, C and C₂, for the "stick drawings", and the particular one chosen was in a convenient place for the 3d drawing showing the projection. By contrast, we pick slices that are quite close to blue/yellow for the lightness/chroma/saturation slice graphics because they best distinguish between different lightness definitions. –jacobolus (t) 00:39, 6 February 2010 (UTC)

I think switching the colors (the input parameters) makes things harder to discern what is happening in the charts (the formulas). You want to keep the number of variables to a minimum when explaining something complex. In fact, I think it would be a good idea to plot the actual colors from the stick drawings in the lightness/chroma charts. SharkD  Talk  00:53, 6 February 2010 (UTC)

Try graphing them: the relationship is much less clear. –jacobolus (t) 01:00, 6 February 2010 (UTC)

I mean something like this (except using the blue and corresponding yellow from the stick drawings). I think the blue and yellow from the stick drawing is sufficiently "close to blue/yellow" to highlight the differences in lighteness. SharkD  Talk  01:47, 6 February 2010 (UTC)

I think that just adds clutter, and doesn’t really help explain anything. I dunno.... maybe it’d be helpful for someone. –jacobolus (t) 02:14, 6 February 2010 (UTC)

Also, no, it wouldn't help for the lightness drawings, because we want to have luma, component average, and HSL lightness be far apart and easily distinguishable. –jacobolus (t) 02:17, 6 February 2010 (UTC)

I think that making the article easier to understand/more cohesive by tying the sections/images together would be better here than slight tweaks to elucidate one particular small piece of minutiae, especially when the changes are marginal. SharkD  Talk  03:19, 6 February 2010 (UTC)

Which “particular small piece of minutiae [sic]” are you talking about? –jacobolus (t) 07:16, 6 February 2010 (UTC)

Also, I think “color space” is more appropriate, as this depends completely on the particular RGB space in use: each RGB color space implies a different HSL/HSV representation. Whereas “RGB color model” to me indicates the overall idea of RGB. –jacobolus (t) 00:21, 6 February 2010 (UTC)

Yes, that's exactly why I think "color model" should be used. HSL and HSV are transformations of the "overall" RGB model. If mapped to a particular color space, their shapes would look different based on the particular gamut used (such as in the sRGB and Adobe RGB images in the 'Disadvantages' section). Whereas a plot of the models' shapes would always remain the same. SharkD  Talk  00:48, 6 February 2010 (UTC)

Huh? I don’t know what “If mapped to a particular color space” means, in this context. If you plotted their colors in, e.g., CIELAB, you’d end up with the same gamut as the plots shown... because it’s the same colors either way, RGB or HSV. –jacobolus (t) 01:02, 6 February 2010 (UTC)

As in, mapping the HSL and HSV color models to a particular color space. If you were to place the coordinates with respect to differences in gamut, then the resulting cylinder would be distorted in much the same way as the RGB cube is in the CIELAB images. SharkD  Talk  01:55, 6 February 2010 (UTC)

That doesn’t make any sense. You’d never try to compare gamuts of two RGB color spaces with respect to the HSV representation of one of them. It would just be a very complicated and arbitrary shape... and either way it has nothing to do with any distinction between “color model” and “color space”. –jacobolus (t) 02:14, 6 February 2010 (UTC)

That's exactly what Munsell talks about in his books: a representation of gamut in a cylindrical (or sometimes spherical) space. And whereas generally labeling things as color spaces would be appropriate in an article about the Munsell system, it would not be here, except in sections specifically discussing their application to color spaces. SharkD  Talk  02:40, 6 February 2010 (UTC)

What? This all has nothing to do with Munsell; he predates the term “color space” by decades. I really have no idea what you mean. –jacobolus (t) 07:16, 6 February 2010 (UTC)

Third opinion on color model, color space, charts

I'm responding to a request for a third opinion:

# Talk:HSL and HSV#break - Two editors disagree on whether the term "color model" should be used instead of "color space" where explicitely [sic] needed, and whether the same color should be used throughout all the charts. 03:13, 9 February 2010 (UTC)

My impression is that "model" properly applies to the overall concept while "space" properly applies to what the model describes, but I'm not an expert on this technical subject.

As to colors in charts, I'm not sure what is meant. Because the charts are intended to convey information about technical aspects of color itself, any color used to format those charts (background, diagrams, etc.) may distract or detract from the technical color content. It seems to me that neutral background colors and non-color lines in a uniform format best support the technical color data in the charts.

WikiProject Color has a banner at the top of this page. If an undertstanding isn't reached here, seeking further input on Wikipedia talk:WikiProject Color might help. – Athaenara ✉ 20:24, 13 February 2010 (UTC)

I don’t think there’s any particular problem coming to agreements on this page. The main question looking for more input, as far as I understand, was the question of the overall page title, whether it should be “HSL and HSV” or rather “HLS and HSV”, since “HLS” seems to be a more popular choice among technical sources. It’s not really such a big deal though. The question of “color in charts” was I think perhaps related to SharkD’s suggestion that I redraw half of these diagrams so that only a narrow set of examples are used throughout, in an effort not to confuse readers. To which my response was that that seems like a lot of extra work for little concrete benefit, and that the particular choices in the diagrams as drawn were careful and not arbitrary. I think I’d add a third response, which is that having *different* examples in this article can also be helpful for readers, insofar as it helps them become fluent with the mathematical abstraction involved. Cheers. –jacobolus (t) 04:17, 14 February 2010 (UTC)

I commented here because of some discussion on WT:3O. SharkD sees this issue as distinct from the HSL/HLS question and wanted more feedback here. – Athaenara ✉ 06:48, 14 February 2010 (UTC)

WRT the use of colors, many of the charts have an "example" color of some type. I think the same color should be used throughout so that readers don't feel lost. The article is continuing to grow more technical in nature, making this more and more likely to happen. The necessary changes would require additional work on jacobolus's part, as he has not shared the source code/project files associated with the images.

I am feeling increasingly hesitant to respond to jacobolus directly, as he is beginning to answer questions with questions, leading to the risk of arguments becoming circular. SharkD  Talk  15:56, 14 February 2010 (UTC)

Well, I suppose “questions” – to my mind I’m asking for clarification because I don’t understand what you mean.... this “color space” / “color model” question, and also the “which particular colors to use in examples” question seem like minor nitpicks, and the arguments for them (e.g. something I couldn’t understand about Munsell, and something else about unspecified minutiae) seem arbitrary/irrelevant. My vague impression is that I haven’t adequately explained some part of what I mean in the article, and SharkD has been left confused about how some part of HSL/HSV work, but it’s a bit hard to tell quite what in this asynchronous chat. I imagine it would be quickly cleared up if we could stand next to a chalkboard for a few minutes... alas. Cheers. –jacobolus (t) 10:06, 15 February 2010 (UTC)

As for “grow more technical in nature” .... as far as I can tell, the article is not especially more technical than it was before, just a bit more precise and much more comprehensive. Previously, it included a few vague and completely unsourced paragraphs which said very little; those have been excised and replaced with some attempt at cited history/comparison sections (which could still be fleshed out a bit, probably). Besides those, it had a big block of pure formulae without a particularly clear explanation of what they meant geometrically. I’ve tried to slow those bits down and fill them in with some prose explaining what’s going on. To my mind, that makes the article more accessible to non-technical readers, rather than less. YMMV I suppose.

I'll try to be specific in my criticisms. here is the old version. The lede is pretty much fine, but the top section “Usage” is vague, unsourced, and grammatically awkward. Its language (e.g. “another visualization of the HSV model is the cone”) is somewhere between misleading and wrong, since HSV is, geometrically, quite plainly a cylinder (see e.g. this article which complains about such usage in Wikipedia and elsewhere). Then the following section “comparison of HSL/HSV” doesn’t really show much of a comparison (more internal slices need to be shown I think for people to really see how the representations differ), and its language is weasely opinion “for some people...” (those some people referred to are mostly misinformed, anyway, IMO). The sections about “Comparison with other models” and “Comparison with color science” are woefully incomplete, and make more weasely statements “it is generally inadvised...”, and the bit about luminosity is not even relevant to color science, being in fact about Adobe Photoshop. Then there’s a bit about older models, which states (sans sources) that these models are “related” to models like Runge’s, and wrongly implies that there’s no difference in concept between saturation and chroma. Finally, we have a “formal specifications” section which basically just dumps a wall of formulae on the reader, and a “software support” section that lists off every image editor any reader of the page had ever used (see WP:NOT). I think that it (i.e. the article as it stood previously) could only be classified as “non-technical” insofar as it didn’t really cover the subject, and readers only “not getting lost” because they weren’t getting much of anywhere. –jacobolus (t) 10:12, 15 February 2010 (UTC)

use "HLS" instead of "HSL"

I've been reading all the papers I can find about these representations, and it seems that "HLS" is used as a name quite a bit more frequently than "HSL", especially in computer graphics literature. Both are very common, and on the web a query for "HLS color" turns up 290k results vs. 240k for "HSL color". So either is probably okay as a name for this article. But we should strive to be as accurate as possible, etc. So unless there are any objections, I'm going to move this article to "HLS and HSV" once I'm finished rewriting it (see previous talk page section). –jacobolus (t) 17:22, 29 January 2010 (UTC)

There's a "Did you mean to search for: HSL color" message at the bottom of the Google search results. SharkD  Talk  00:34, 31 January 2010 (UTC)

Also, when searching for both search terms, "HLS HSV color" results in 135k hits, while "HSL HSV color" results in 985k hits. SharkD  Talk  00:38, 31 January 2010 (UTC)

I think that HSL is more common now than it once was (probably thanks at least in part to Wikipedia!), and especially when both are described people are tempted to mix up the letter order so they’re consistent, but HLS is used in almost all of the scientific papers about the subject, and by other “careful” sources (for instance, the Mac OS 8.0 color chooser which has both HLS and HSV modes, presenters at SIGGRAPH, etc.). I don’t think it’s too big a deal, but Wikipedia should strive to be as accurate as possible. –jacobolus (t) 02:05, 1 February 2010 (UTC)

I guess. Here's a breakdown:

Edited several times since the initial posting.

Search term		"HSL HSV"	"HLS HSV"		"HSL color"	"HLS color"		"HSL colour"	"HLS colour"		"HSL adobe"	"HLS adobe"
Google Web		1160k	279k		245k	278k		3160k	343k		92k	132k
Google Blogs		1856	1166		9435	8680		10690	8680		4370	3422
Google Books		627	636		758	834		804	926		420	313
Google Scholar		1050	1390		6590	11800		6360	20100		742	7690
Google Web (2000 - 2010)		unavailable
Google Blogs (2000 - 2010)		1809	1153		9229	7585		10473	7884		4276	3062
Google Books (2000 - 2010)		305	332		666	650		674	667		309	245
Google Scholar (2000 - 2010)		524	681		3150	4150		3070	16800		450	3720

I don't see "almost all of the scientific papers" anywhere in these numbers, or the relevance of the Mac OS color picker as the Windows color picker uses HSL. As far as book authors go, there's not much of a difference in numbers or date of authorship. Not sure what policy is in this regard. Generally, we're supposed to go by what most people are familiar with, for instance "dog" versus "Canis lupus" or "Zoe Saldana" versus "Zoë Saldaña". SharkD  Talk  03:12, 2 February 2010 (UTC)

w/r/t date of authorship, the original dimensions in Joblove and Greenberg (1978) were called "hue", "intensity", and "relative chroma". Then Tektronix introduced the 4027 terminal in 1979, which called it HLS (the only difference from the version described in this article being that blue was placed at 0°). Also in 1979, the Computer Graphics Standards Committee recommended the space, and called it HLS. Following that, every source from the 1970s and early 80s calls it HLS. By the time we get to the 1990s, both terms are in use, but the most authoritative computer graphics textbooks, and papers presented at SIGGRAPH, etc., seem to still almost exclusively use the term HLS. Microsoft’s color picker doesn’t use an abbreviation for this space anywhere, as far as I know: they just label their input boxes. Photoshop also doesn't put the abbreviation anywhere: it just has a "hue/saturation" tool. The real notable number in the chart above is the "HLS color" number from Google Scholar. –jacobolus (t) 03:42, 3 February 2010 (UTC)

While it would be good to mention it for a historical perspective, I don't think it's necessary to change it in the text. Lots of terminology changes over time. Science and technology aren't static. My remark WRT the dates of authorship was in response to your statement that Wikipedia had somehow influenced the term's presence in external texts. The dates are spread fairly well across a wide time frame, leaving this conclusion in doubt. SharkD  Talk  10:14, 3 February 2010 (UTC)

I added a few more rows to the table with additional statistics based on date. Yes, the gap is growing closer, but HLS still seems to be in the lead as far as scholarly works and books are concerned. (I don't have data for Google Web.) As CSS3 takes hold this trend will probably continue. I also requested assistence over at WP:Village pump (policy)/Archive 71#Popular and recent usage, or scholarly and historical usage?, and they should get back to us here. SharkD  Talk  10:47, 3 February 2010 (UTC)

Also, I'm pretty sure that we call her "Zoe Saldana" because that's her name according to the SAG, how she's listed in movie credits, etc. –jacobolus (t) 03:46, 3 February 2010 (UTC)

Yes, that's my point. It's how we all know her now, not what she was (or may have been) named in the past. (The sources aren't consistent.) SharkD  Talk  10:14, 3 February 2010 (UTC)

One last bit of confusion: Alvy Ray Smith's 1978 paper describes HSV and a "hue"/"saturation"/"brightness" (HSL) model, but what he calls the HSL "triangle model" is quite different from the HSL/HLS that we are describing in this article. –jacobolus (t) 04:15, 3 February 2010 (UTC)

I would like to see the paper if you happen to have a copy. SharkD  Talk  10:14, 3 February 2010 (UTC)

I can’t find your email address on your user page or homepage. Shoot me an email and I'll happily pass along any of these papers you want to see. –jacobolus (t) 04:40, 4 February 2010 (UTC)

Firstly, don't forget "HSL colour" and "HLS colour". Personally, I have studied computer graphics courses and colour models at masters level, and worked with them for the past year, and have never seen "HLS". OrangeDog (τ • ε) 12:38, 3 February 2010 (UTC)

OK, I added a column for "HSL colour" and "HLS colour". The Web vs. Scholar schism seems even more pronounced. SharkD  Talk  05:52, 4 February 2010 (UTC)

I think what might have happened is people re-constructed an acronym from looking at the order of Photoshop's hue/saturation tool ("hue", "saturation", "lightness") or MS's color picker ("hue", "sat", "lum"), or from taking the order from HSV (whose use is more common than HLS/HSL in software), rather than from the technical/academic descriptions, and then because those web descriptions were more accessible than journal articles, other people just looking the thing up came more and more to use the "HSL" order, to the point where among all the non-technical vague descriptions that one finds littered around the web, "HSL" has even overtaken "HLS" as the most common order. Not sure where that leaves Wikipedia though... I still think we should go for the abbreviation used more often by experts. –jacobolus (t) 10:59, 4 February 2010 (UTC)

I asked for a third opinion since the village pump didn't result in much comment. SharkD  Talk  03:37, 5 February 2010 (UTC)

Response to third opinion request:

Is there any reason we can't simply allow both names to exist by leaving this page named with HSL and creating a redirect from the HLS page. The empirical search results above seem to indicate that (at least among search engines) there is no clear correct name for this. The article already contains a helpful reference to the fact that "HSL is often also called HLS or HSI, and HSV is often also called HSB". I am going to leave the 3O request open because this could probably benefit from more than just my eyes on it.  7  03:46, 5 February 2010 (UTC)

.

Response to third opinion request:

I also believe that both names should be left. After all it seems clear from above that there is no proper or "official" term. It would help to have both. Perhaps stating which one is most commonly use or recognized. 04:02, 5 February 2010 (UTC)—SoCal L.A. (talk) 04:02, 5 February 2010 (UTC)

It affects the name of the article as well. We already have two acronyms in the title. Adding the full selection would be a huge mess. Personally, I am more familiar with HSL, and like the symmetry with respect to HSV. SharkD  Talk  04:40, 5 February 2010 (UTC)