Talk:Color space

From Wikipedia, the free encyclopedia

This article is supported by the Color WikiProject, 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 This article has been rated as start-Class on the quality scale.
High This article has been rated as high-importance on the importance scale.

Contents

[edit] Usefulness

This article is fairly technical. I think the introduction should be expanded with a brief explanation of how color spaces are actually used by artists and graphic designers today, i.e. what problems they solve. In particular, laypersons are now exposed to color space terminology via popular PC software such as Adobe Photoshop, and these readers are probably more interested in "Why do we need it?" than "How was it developed?". 151.199.131.100 11:47, 15 July 2006 (UTC)

[edit] Untitled Comments

I'm not sure what to do about this, but I think it important to mention that there are older systems, particularly the Mussell (Hue, Value, Chroma) system which predates nearly all of the other systems and, IIRC, was designed for painters to be able to talk more accurately about the paint colors they wanted.

LAB, is another popular color space, which should be mentioned. I used to work on a research project to match dental porcelain colors to actual tooth colors, and I had to write a program to calculate color differences. The LAB system is perceptually equalized so that the color differences are quantifiable, but it is nearly impossible to think in the LAB system. On the other hand it's pretty easy to visualize the proper color once you know it's Hue, Value and Chroma, so products are generally sold by Munsell color, but the Munsell system is not perceptually equalized because the color difference between one hew and another is arbitrary, so I had to convert everything over to LAB before any calculations could be done.

My point here, is that there is lots of important information about Color Spaces which has nothing to do with RGB, or CMYK. In particular, I think the notion of regular and irregular color spaces ought to be included somehow. On the other hand I did this work a long time ago (when I was 16) so my memory is a little fuzzy, and I have no references immediately available on the subject. MRC

Actually, the hues in Munsell are more perceptually uniformly spaced than in L*a*b*. It's just that if you go in an arbitrary direction in Munsell (varying hue, value, and chroma all at once), the distance between colors doesn't particularly imply their perceptual difference. Unfortunately, human color vision is non-Euclidean, so it is impossible to have a three-dimensional color space for which distance between points in that space accurately quantifies color differences. But yes, this article is very lacking: it's missing any kind of historical discussion, and it completely misses many of the most useful color spaces. --jacobolus (t) 06:46, 17 May 2007 (UTC)

From the article:

Note: this article needs expert input: at the moment, it is extremely informal, and needs lots of work by experts in the field to give a good account of this highly technical field.

From the article:

"For example, although several specific color spaces are based on the RGB model, there is no such thing as the RGB color space."
In 1931, Commission internationale de l'éclairage, or CIE, developed a Red-Green-Blue (RGB) model of human color vision. Their results were combined into the specifications of the "CIE RGB color space". -Dawn McGatney 69.139.231.9 (talk) 17:24, 18 March 2008 (UTC)

I've been working a bit on the HSV color space page. Maveric brought up an interesting point; should we be making the distinction between a color model and a color space? As I understand it, (someone correct me if I'm wrong) the color space is the region of possible colors that is described by a particular color model. Color models would include RGB, CMYK, and so on, but sometimes, two different color models will describe the same color space (such as HSV and RGB). Or would the term gamut be more appropriate to describe the collection of possible colors defined by a color space?

I found colors can be represented in Stokes vector.

Is there any identical color model to this Stokes vector form?

The parameters of HSB are same that of partialy polalized light.

  • Hue is the azimuth of linearly polarized light.
  • Satuation is the degree of polarization.
  • Brightness is the total intensity of the light.

Hue is represented in [0 - PI] in Stokes vector, not [0 - 2PI].

Stokes vector form of color is very useful for color transform. For example, hue rotation is performed by just a matrix product. -- goto

Reviving the old question above: We should really try to be clear, in all color-related articles, on the difference between color space and color model, if indeed there is such a distinction. The terms seem to be used interchangeably; the article as it currently stands makes a fairly good distinction between them, but confusion could result, especially since the article begins with a definition of color model, rather than color space (and even more confusingly, because color model [used to] redirect to color space).

Additionally, there are problems such as Lab color space opening with "CIE L*a*b (CIELAB) is the most complete color model used conventionally...", odd because it does not say "The Lab color space is..." and because it does say "...is [a] color model." And of course, there are these:

  • RGB color model which strangely does not link to RGB color space, a separate article
  • HSV color space which has no corresponding "model" article, and opens with "The Hue Saturation Value (or HSV) model..."
  • CMYK and YUV (no "space" or "model", and YUV is described as "...a color space in which the Y stands for the luminance component..."---though, if we're going to distinguish "space" from "model", a color space has no Y or any other component, though a color model does)

Etc. Anyway, this could probably use some clarification across all the color articles. -- Wapcaplet 19:42, 5 Mar 2004 (UTC)

Man, you're more than welcome to kick some life into these articles! I was checking them out recently, and after finding a lot more inconsistencies than the ones you outline above, I started polishing them. All my edits do however acknowledge the fact that more work needs to be done on those topics, as you can see from my summaries. I just didn't have the guts to make really radical changes in the articles, although I did my share (e.g. separating RGB color space from RGB color model, along with some radical edits within gamut and color space). --Gutza 00:45, 6 Mar 2004 (UTC)

I like the work you've done to this article - bravo! I'm not so sure I like the idea of having a separate model/space article for each model/space. Though, having either one redirect to the other could just cause more confusion, so maybe it is good to keep 'em separate. It just doesn't seem like there'd be much to say about XYZ color space, aside from the fact that it's defined by the XYZ color model. It's about time we had an article on gamut though (which is looking really good!). Dunno if I'm ready to dive in and clean all these up yet either... but hopefully we can get the ball rolling. -- Wapcaplet 01:13, 6 Mar 2004 (UTC)

Ok, don't want to sound rude (why does that syntagm always induce the feeling that something rude follows?), but have you had the curiosity to diff those articles to the state before my changes? I mean, I agree with most of the points you criticise (e.g. XYZ color space irrelevance, color space/color model redirects), but the respective articles had the respective faults and were also in a much worse shape before I edited them today. That's why I feel obliged to reiterate: if you feel like diving into radical edits, by all means, be my guest! These articles stayed in an awful shape for months, so I can only feel good for raising the flag.
The only thing I don't agree with in your comments is the fact that you seem not to agree with the idea of "separate model/space article[s] for each model/space". Given the particularities of each model and each space, I don't think it's out of line to have distinct articles on each of the mainstream color models/spaces. IMHO that can also be useful for other authors as well: an author who wants to refer specifically to sRGB for instance would be happy to be able to link to the sRGB article directly instead of explaining what sRGB is within an otherwise unrelated article. --Gutza 02:09, 6 Mar 2004 (UTC)

I think I may have been unclear; what I meant was that I am not sure I see the benefit of having both RGB color space and RGB color model, for instance (at the risk of contradicting my earlier statement that the two concepts should be made distinct). Certainly, I agree that each color system---RGB, CMYK, HSV, etc.---should have its own article, as there is a lot to say about each. I fear that having both RGB color space and RGB color model risks a lot of unnecessary duplication, since the two subjects are so closely related; perhaps a title such as RGB color system or just plain RGB might be better, as it avoids the problem with "space" and "model" being subtly different; the distinction between "space" and "model" can then be made in (I hope) a more neutrally-titled article, if such a distinction is necessary for that article.

Whether there should be a separate color model article, I don't know. I'm fairly sure you know more about it than I do, so your input on that would be good. -- Wapcaplet 02:50, 6 Mar 2004 (UTC)

The reason why I separated RGB color model from RGB color space is the fact that there are several color spaces based on the RGB model, but there's only one RGB model. (Please note that RGB and RGBA point to the RGB color model article, because that's most probably what something referring to RGB means.)
Another reason is that RGB is the color space champion, everybody uses RGB color spaces on a daily basis, we just don't realize it (when scanning documents, printing or editing images, or even surfing the web). Also, everybody's buying digital cameras these days, people will want to know more and more about these topics. That's why I think it's important to make that distinction.
With CMYK things are even more complicated: there are probably even more color spaces for CMYK, but people don't "feel" those as color spaces, but rather as "profiles" used to describe the output device.
Anyway, what I'd like to end up with would be the RGB color model and CMYK color model to be the "article proper" explaining the actual model and generic formats, while RGB color space and CMYK color space would be something similar to what RGB color space is right now: just a short note on the fact that "an RGB color space is any color space based on the RGB color model", along with a list of the most proeminent color spaces in use (sRGB and Adobe RGB for instance). That way the coordinates are established, links to the relevant articles are provided, and anyone interested in the topic can read more about it in those respective articles.
The question whether color space should be distinct from color model. I don't really know, I can only give an opinion. If we were to treat the problem rigurously encyclopedical, they should be distinct articles. They should only be distinct however in a highly specialized encyclopedia where we would assume that all readers are graphic designers familiar with the concepts in the first place. If the target is the layman, which I think is the case in a generic encyclopedia, then I think the right way to go is combining the two in one article explaining the differences (much like how it is now). --Gutza 14:02, 6 Mar 2004 (UTC)

Valid points. I never really considered that there are many different RGB and CMYK color spaces. I suppose the same would be true of other color spaces as well; would it be true to say that a color space is influenced not only by the corresponding color model, but also by the environmental conditions (lighting) and properties of the output device (printing inks, monitor phosphors)?

I agree with having RGB color model as a main article, and RGB color space as a more detailed discussion of RGB color spaces. For most color models, it won't be necessary (not as of now, at least) to have a separate color space article, so those should be unified into the main HSV/YUV/CIE color model articles until there's a need to split them off. I agree that color space should remain a single article, unless there becomes a strong need to have a separate color model article. (Or should the main article be color model instead of color space?)

We are making progress, I think :-) -- Wapcaplet 16:19, 6 Mar 2004 (UTC)

You're absolutely right throughout the posting above. To get it out of the way, yes, I also feel that the HSV/YUV/Lab models don't really need to be split. I might be wrong though, I haven't read much literature on these because I personally never use them in order to avoid converting my images back and forth and degrading them unnecessarily.
Regarding your first question, what you're describing is a device profile. In an academic sense, yes, that could also be called a color space. But it's not really what's usually meant by color space in common use. Let's clarify sRGB vs. Adobe RGB, which are the color spaces all digital camera owners will want to know about, so you can get a better idea. (I'm talking chiefly about Adobe RGB 1998 here, that makes the difference clearer.)
  • sRGB is a color space used to standardize generic monitors. It's basically the de facto RGB model default color space. It's used primarily for web design, that's what you'll read in most FAQ's out there, but should be used generically for all processes targeted at computer displays (interactive CD-ROMs, computer games, screen-savers, kiosk applications, whatever gets to be shown on a computer monitor).
  • Adobe RGB is a much wider color space. It has a huge gamut, and it's basically used to encode a much wider variation of colors in the same 24 bit structure as sRGB. There are two things which need to be pointed out regarding Adobe RGB:
    1. The color space is much wider, but the number of colors which can be encoded in 24 bits is the same. The result is a wider distance between colors. The gamut is "thinner" if you wish, it's less dense than sRGB. The problem with this is that further manipulating the image will result in degrading it sooner than if you were to use sRGB (see the RGB article for a little more on this).
    2. Not all monitors are able to display the whole Adobe RGB gamut. You can wind up using Adobe RGB just to find that you messed up your images in processing because you have the wrong feedback about what you're doing. Apple used to make some great monitors with a gamut about twice the size of a regular monitor's (I don't know if they still do, but I would expect them to).
I hope that makes it a little clearer. A color space (or "working color space", as they often call it) is a way to encode colors from the Lab space into the RGB model. On the other hand, you should also calibrate your input and output devices (scanner, digital camera, monitor, printer) and produce profiles for those devices in order to properly reproduce the original color space (for input devices) or desired color space (for output devices).
Now there's a trick to this. When converting an image to a different color model along with the color space, it's great if the target color space is the actual color space of the target device. That avoids converting to the target color model in one step, only to convert again to the proper color space within the respective color model in a subsequent step. That's great because every such conversion further damages the image. As a result, when converting your RGB image from whatever color space you were using to CMYK for print, it's recommended that you convert directly to the color space of the target device in one go. So basically you're incorporating the device profile in the target color space definition.
That's why technical documents generally say that the CMYK color spaces are device-dependent, as opposed to the RGB spaces which are device-independent. I don't care what you're displaying my Adobe RGB image on, just make sure you calibrate that thing to show the proper colors. That's because my image along with the color space information map the RGB triplets to "real world colors" within the Lab space. But if you want to print my image, tell me what you want to print it on, so I can convert it directly to your printer's proper CMYK space.
The terminology is pretty ambigous, I know, sounds like sometimes two things are the same, while other times they're distinct. I can only hope this posting sheds a little more light on hot the terms are generally used. Because academically color spaces and device profiles really are the same thing. --Gutza 18:04, 6 Mar 2004 (UTC)

The following paragraph is too RGB-specific for this general article, I think.

The RGB color model can be implemented in various specific incarnations, depending on the capabilities of the system used. By far the most common general use incarnation at the time of this writing is the 24 bit implementation, with 8 bits per channel. That means that each of the three channels in such an RGB implementation can store 256 discrete levels of color. Any color space based on the RGB model is thus limited to a gamut of 256×256×256=16.7 million colors. Therefore a "small", "restrictive" color space based on this RGB implementation will result in exactly the same number of colors as a "large", "generous" space. The only difference between the two is the volume their gamut cover, but that is always inverse proportional to its density on the reference space.

Smyth 12:23, 31 Aug 2004 (UTC)

No, it's not. It does use the means of the RGB example to prove a point, but that's only because RGB is the model understood by most people. If CMYK or some other bit depth would've been the most widespread, then I would've used the CMYK example with the respective depth to prove the same point: color space density over Lab space. I'll copy it back in the article.
Good job on the rest of the restructuring though, great copyediting!
-Gutza 12:52, 31 Aug 2004 (UTC)
Thanks, it was fun.
Smyth 13:08, 31 Aug 2004 (UTC)

I hope I haven't clashed with any work in progress by adding to the sRGB and Pantone artices, and adding every color space I could find to the Color space category. I didn't see this discussion until just now. Also: is there a discussion of ICC profiles anywhere? Notinasnaid 16:21, 31 Aug 2004 (UTC)

[edit] 16-bit color channels: accuracy

"An implementation which uses 16 bits per component will result in exactly the same range of colors as the 8 bit version, but to a 256×256×256 ≈ 16.7 million times higher accuracy."

This is surely wrong. Is 16 bit/channel RGB 16 million times more accurate than 16 bit/channel greyscale, and is 16/channel bit CMYK 256 times more accurate than 16 bit RGB? No, surely not. On the other hand this replaced a section that said "twice as much" accuracy, and that's clearly wrong too. (Analogy: if you double the number of decimal places you do much more than double the accuracy of a number).

It might be safest just to say "substantially higher accuracy". Thoughts? Notinasnaid 09:23, 4 Feb 2005 (UTC)

  • I think "accuracy" is the wrong word to use. Accuracy refers to the precision with which some original reference point is measured. If the 16-bit image in question is a scan of a photograph, then the color accuracy would be how closely each pixel matches the color in the photo (under some kind of ideal lighting conditions). For a purely digital image, the term doesn't even apply. I would think accuracy should be measured as a percentage - a digital image is somewhere between 0% and 100% accurate (at least in terms of the colors represented). An increase in bits-per-channel obviously increases the number of distinct colors, but I think it's important to qualify that fact with a mention (preferably with some references) that humans can't distinguish that many, and most display hardware can't display that many. AFAIK, the only reasonable application for 16-bit/channel color is as an intermediate format for filtering, compositing, and other manipulation (to preserve as much existing "accuracy" as possible). Some mention is made of this now. I will rephrase. -- Wapcaplet 16:51, 4 Feb 2005 (UTC)

Much better. On reflection I think the original (not gone) phrasing was seeking 'precision' rather than 'accuracy'; and the precision is 256 times. Notinasnaid 22:09, 4 Feb 2005 (UTC)


It was said "It just doesn't seem like there'd be much to say about XYZ color space, aside from the fact that it's defined by the XYZ color model." Personally, I think it is very important that the CIE XYZ space is mentioned, as correct color space conversions are expressed as functions of XYZ coordinates of the spaces participating in the translation. I feel the following points are absent from the current article: - linearity of color spaces, with proper notation ("prime" notation for non-linear spaces, such as Y'PbPr, and R'G'B' vs linear RGB) - coordinates/definition of a color space (color primaries, white point) - quantization and gamma correction, along with most commonly used gamma curves - add more examples of color spaces - scRGB, explain "studio RGB" vs "computer RGB" - clarify which Y'PbPr/Y'CbCr spaces are being referenced - ITU-R BT 601, BT 709, SMPTE 170M etc etc.

Lastly, it should be pointed out that the R'G'B' spaces currently used in computer graphics/video are, in fact, represented on 32 bits. Indeed, 24 bits are actually used for color channels, but ease of addressing and manipulation dictated a full double word is used. The remaining 8 bits are used for storing transparency information (alpha). 10 bit formats are emerging (A2R10B10G10).

I'm somewhat of a newbie to editing Wiki, and so I didn't think it was proper to proceed with these modifications prior to bringing them to attention in the discussion page. 08:11, 12 April 2006 (UTC) D.

[edit] 3D

A color space is multi-dimensional. A RGB color can thus be seen as a 3D object: Red=X, Green=Y, Blue=Z.

It needs to be pointed out that RGB, LAB, YIQ and HSV can be seen as 3D objects. An illustration of a 3D view of a colors space would make the concept much easier to understand! --Kasper Hviid 12:14, 26 Feb 2005 (UTC)

Image:HSLSphere new.jpg
64px
Here are two. I would be interesting in knowing how to model the other color spaces. -SharkD 08:09, 23 October 2006 (UTC)

[edit] Let's remove Color Model

I am unsure whether the article always uses "Color Model" as a word in itself, or whether its sometimes uses it as a synonymous for "Color Space".

In any case, the Color Model part needs to be stripped into a article on its own. However, I don't feel I have the knowledge to do this.

If I may quote from color management for photographers:

A color model is a method of grouping numeric values by a set of primaries. Most color models have three primary components (e.g., RGB, CMY, LCH, HSV, L*a*b*). Some application-specific color models use more components, for example, CMYK. Photoshop and other image editing applications work with multiple color models. Some scientific color models such as x,y,Y or L*a*b* encompass all of human vision and have a defined scale such that a particular color will always have the same set of values. It’s quite useful to be able to assign a numeric value to a color based upon how humans perceive colors. Other color models such as RGB or CMYK have no standard defined scale or reference. These color models are abstractions and cannot describe a specific color with out first defining the scale or reference. Having an RGB value such as R10/G30/B50 does not tell us how to reproduce that color; the values provided are simply a ratio of the primary components (in this case RGB) without a necessary scale. In this example we know there’s more blue than red, more green than red, but not as much green as blue. Without a scale to tell us the actual amounts we can’t reproduce this color. This is where a color space comes into play. A color space provides this additional and necessary information.

Think for a minute about the term color space. The first word, color, is fairly obvious. Let’s look at the term space in context. I have three sets of numbers to define red, green, and blue (R10/G130/B50) but how red or green or blue are these values? What is the scale? It is possible to plot any three primary values in three dimensions by treating the primaries as coordinates in space. This creates a solid or volume that represents all possible colors in that color space. This is a color space! The color space exists within the larger universe of human vision. Each color space is in a different position relative to this universe. Thus the same RGB values in each color space will be in a different location.

Going back to the chocolate chip cookie analogy, suppose a color model is a cookie recipe with only three ingredients. I give you this recipe, which simply calls for 1-flour, 8-butter and 2-chocolate chips. You don’t have enough information to make the cookies. However if I provide you the recipe with a specific scale—1 cup of flour, 8 tablespoons of butter, and 2 cups of chocolate chips—I’ve provided the necessary information, the scale, to make a dozen chocolate chip cookies. If I provide additional information such as the brand of chocolate chips, you can reproduce exactly the same cookies I made. I can give you the cookie recipe in the metric scale such as liters and grams and you can still makes the same cookies even though the numbers are different. A color space is a color model that has a known reference and scale, in this case primaries (the ingredients) and scale (specific quantities of these ingredients).

I'm not sure if I agree with everything above, but it seems to contradict the third paragraph of the introduction. I don't believe that a bunch of color swatches constitutes a color space. An unsophisticated color model, yes, but color space, no. PAR 02:01, 16 July 2006 (UTC)
I haven't looked into this in detail, but Wikipedia includes both RGB color space and RGB color model articles. Certainly, people do get confused between them. However, as color spaces often dervive from color models, this article at least needs to note that. It may say too much; there is no overall coordination between articles, so there is more repetition than is desirable. This can only be resolved by having an active coordinating project, not piecemeal. What seems to have happened in this article is that the author(s) of the opening paragraphs haven't followed the Wikipedia standard to define the term in the opening sentence, and have in fact decided to define a different term. Notinasnaid 09:09, 17 July 2006 (UTC)
It seems to me that the color space article spends most of its time talking about color models. and the color model article spends most of its time talking about color spaces. -SharkD 15:48, 26 October 2006 (UTC)

[edit] Image too dark

The image for subtractive color mixing seems too dark. Is this intentional? Otherwise, we could just make a negative of the additive mixing image and use that instead. -SharkD 08:11, 23 October 2006 (UTC)

I think that it is reflecting the reality of CMYK colors, which are nothing like the complements of RGB colors. Notinasnaid 10:42, 23 October 2006 (UTC)
The other, side-by-side gradient image is a better comparison of the CMYK and RGB models. Having the color mixing images be from different palettes might be misleading. -SharkD 06:44, 27 October 2006 (UTC)
It's not a particularly good comparison, because many saturated (particularly dark) CMYK colors cannot be displayed on a computer display. --jacobolus (t) 06:49, 17 May 2007 (UTC)

[edit] ISCC-NBS

Why is there absolutele no mention of the ISCC-NBS color model? --Metallurgist 00:41, 15 November 2006 (UTC)

[edit] Cone Contrast Space

From the article:

However, a color model with no associated mapping function to an absolute color space is a more or less arbitrary color system with little connection to the requirements of any given application.

Cone Contrast Space if defined by the change in response of the three cone types. The origin of this space (0,0,0) is defined by the current state of adaptation of the three cone types, so it is certainly not absolute or arbitrary. And it is the most useful space to use if you want to model the visual system. Drichters 02:23, 23 February 2007 (UTC)

[edit] Remission?

I have reverted the edit changing "reflected" to "remitted." While "remission" may be a term of art in spectroscopy, it is not generally used in color science. For example, the term "remission" does not appear in the index of Wyszecki and Stiles's second edition, a definitive reference in color science. Lovibond 16:22, 25 June 2007 (UTC)

I think it was me who changed that part. Per definition (at least as i know it) reflected light does not 'interact' (for lack of a better term - i'm not native english) with the object reflecting it. Remitted light on the other hand does, which is why you can see the object's color. Only mirrors really reflect light (photons?).
Perhaps according to some definitions, but not in color science. Also, please be sure to sign your edits on this page with four tildes, thus: ~~~~ -- thanks! Lovibond 15:23, 10 July 2007 (UTC)

[edit] Merge with List of color spaces and their uses

I would just get rid of the list article and use this navigation template in all color space articles. It needs to be expanded, of course, but that's easy. I prefer navigation templates to "list of..." articles unless the meaning of each element in the list needs to be explained.--Adoniscik (talk) 04:07, 21 January 2008 (UTC)