Talk:Color balance

Untitled
add new talk topics at the bottom of this page

Original version of this article, for reference
In graphics, color balance is the term for the dynamic range of the colors in an image. In artworks in general, the color palette can be selected by the artist to convey a mood or an impression.

In image processing, color balance  refers to the mapping of the distribution of colors in each pixel of the image to lie within the 8-bit range of human perception (256 colors in any given image). Although the visual system of a human being can distinguish many more colors than a camera or a photograph, the subtleties can be satisfactorily simulated by artful color balance of the palette in any given image.

Some implementations of color balance use histogram equalization.

Is any of that correct?
I believe this article suffers from continuing to contain someone's initial unsourced guesses as to what color balance is. Based on my own experience in color science, I'd say nothing above is correct, and not much that has been since then has helped much. I recommend we look for good sources, and replace the nonsense with what they say. Just checking here first to see if anyone thinks any of that is correct, and if so, whether a source is known. Dicklyon 03:53, 6 February 2007 (UTC)


 * No, the above information seems pretty completely wrong, as far as I can tell, although, I'm not sure that color balance should be explained by “relative amounts of red, green, and blue primaries.” That's a pretty arbitrary definition, given the arbitrary nature of those “primaries.”  Instead it could perhaps be described in reference to trichromatic color vision and the opponent process.  And some explanation of adaptation is required, I think.  --jacobolus (t) 11:08, 9 May 2007 (UTC)

Merge “white point” into this article
It seems to me that "white point" and color balance are close enough subjects that there isn't really a need for multiple articles on the subject. We can have subsections about "white point" and "black point" on this page, if need be.

I just sent "white balance" to redirect here, instead of to "color temperature," where it redirected before. “White balance” a.k.a. “color balance” is a bit different from “color temperature”, in that temperature is a one-dimensional value, whereas balance can also be shifted towards green or magenta.

In this article it would be useful to therefore have subsections explaining what "white point," "black point," "color temperature," "color balance," etc. are, and their relation to each-other, and it would be good to explain about human visual adaptation to color balance. I don't know all that much about the subject, but surely there are papers online explaining its properties and limits.

This article as it currently stands is not particularly scientific (I'm not sure why it isn't tagged as a stub); but I think it could be made into a reasonable article, starting with the influx of information from the white point article.

--jacobolus (t) 10:45, 9 May 2007 (UTC)


 * Neutral for now -- seems like a good idea, maybe, but I'm going to have to look for sources and see if they have compatible definitions of these concepts. These are concepts that I've worked with professionally for years, but like many things that one works with, finding verifiable definitions and info is not always easy. Dicklyon 03:55, 10 May 2007 (UTC)


 * ''(moved from talk:white point)
 * I don't support such a merge. These are two distinct but related topics, and are different enough to each merit their own articles. Color Balancing, once the white point is known, is not a fait acompli, and it has been demonstrated that color balancing may be more effective using a separate characterization for each taking/viewing illuminant pair. (See, e.g., "A Comparison of Different White Balancing Algorithms as Quantified by their Color Constancy," Proceedings of the SPIE, volume 5301, 2004.) Lovibond 02:20, 12 May 2007 (UTC)


 * Well, first, you should leave the conversation at the talk:color balance page. But my reasoning isn't really that the concepts are identical, but that they are so closely related that most of the article content would be mostly repeated, and it therefore is more useful to readers to have one article which discusses both, instead of having separate articles.  --jacobolus (t) 05:43, 12 May 2007 (UTC)


 * (moving some more discussion)
 * I'm not participating in any discussion on talk:color balance, so I can't exit that discussion. Fact is, the two topics are quite different. For example, white point plays a key role in the computation of color coordinates, such as L*, a*, and b*, which have little to do with color balance. Color balance can be performed without knowledge of the white point. I believe that merging the two articles would marginalize at least one of the concepts and foster confusion of the two ideas. Merging the two might have a small advantage to people (i.e., those performing gray balancing of digital color imagery) who have a fairly narrow view of the these two topics, but retaining cross-referenced articles would, IMO, enable a much broader view with minimal fuss. Isn't a broader view better for this forum? Lovibond 15:45, 14 May 2007 (UTC)


 * Er… "leave" as in "put", not as in "exit" --jacobolus (t) 21:56, 14 May 2007 (UTC)


 * Oppose. Having considered it more, and listening to Lovibond's points, I agree they should NOT be merged. Dicklyon 22:01, 14 May 2007 (UTC)


 * Okay, I'm going to remove the merge templates then :) --jacobolus (t) 22:28, 15 May 2007 (UTC)

retinex
"retinex" redirects to "color constancy." Should it? Both terms are linked from this article and some others. --jacobolus (t) 00:22, 16 May 2007 (UTC)

Retinex was Edwin Land's name for his color perception theory, which had a lot to do with color constancy. I don't know of any cameras that use a retinex-like approach to auto-white-balance, so I don't think it's very relevant here. Not that it couldn't be done, but it's way more complex than any color-balanceing approach.

The Viggiano paper
Lovibond, this paper is very interesting, but not as good or thorough, in my opinion, as this one (Chromatic Adaptation Performance of Different RGB Sensors, by Sabine Süsstrunk, Jack Holm, and Graham D. Finlayson). They compare more spaces designed for, and commonly used for, illuminant adaptation, including the Bradford and ROMM spaces, both of which work pretty well.

I don't care for what Viggiano did about Camera RGB, since his relatively narrow-band gaussian model of camera sensitivities biases the results in favor of that space; these results do not generalize well to lots of real camera sensitivity sets, and especially not to broad ones like the Foveon sensor's. And your interpretation that "This means that it is advantageous to get color balance right at the time an image is captured" is not supportable in the context of raw capture (the next clause "rather than edit later on a monitor" makes some sense, but only as a strawman to be knocked down, since that's not the real alternative). Besides, the result does not imply that monitor RGB can't be corrected, just that simple balancing is not usually best.

Thanks for your work on this. I'll help as I can find time. By the way, your photos were a good idea, but they got speedily deleted for lack of a proper license.

Dicklyon 04:11, 16 May 2007 (UTC)


 * Süsstrunk, et alli, covers chromatic adaptation, which is a phenomenon of the human visual system, not color balancing. Both are IMO important topics, but one is more germane here than the other. You state that you don't like Viggiano's choice of sensors because they weren't representative of those used in digital cameras. But the sensors used by Süsstrunk, et alli, were the human cone fundamentals, which are even less representative of camera spectral sensitivities. Not a criticism of Süsstrunk, et alli, because they weren't addressing color balance, their topic was chromatic adaptation.


 * The topic is not really that different, except that Viggiano is measuring an objective constancy, while Süsstrunk is looking at human preference. Either way, the question is to find what space is good for doing white balancing by using diagonal matrix gains.  Süsstrunk focused on color spaces that were designed or promoted for this application, which Viggiano focused on the idea of using the camera space itself, but with no actual data to represent what a camera space might look like; there was no support for the idea of the gaussians, but I understand that they do stand in for sensor sensitivities; just not very general, so the conclusion has to be taken with a big grain of salt. Dicklyon 23:04, 18 May 2007 (UTC)


 * Also, Gaussian sensitivities seem to be a surrogate of choice when actual sensitivities are unknown, particularly at RIT, which is where I am. Quan's dissertation is a good example. I'll try to get a reference to an abstracted version. Lovibond 15:58, 18 May 2007 (UTC)

Name of article
Furthermore, I'm not so sure that what we're talking about here is "color balance". It's actually more like what's called "white point adaptation" or "illuminant adaption" or "white balancing". The difference, I think, is that color balance is generally conceived of as a simple operation in a working RGB space or something like that, while the illuminant adaptation usually involves transformations between spaces, and/or custom matrices for different illuminants. The Viggiano paper does not mention the term "color balance" at all. So I'm wondering what should be the boundaries of these topics, or whether we need to call this article "white balance" if that's what it's about.

Dicklyon 04:19, 16 May 2007 (UTC)


 * "Adaptation" is a phenomenon of the human visual system. "Color Balance" pertains to image reproduction. The latter is also referred to variously gray balance, white balance, etc. Lovibond 15:58, 18 May 2007 (UTC)


 * But adaptation to the illuminant is also something that a camera must do to keep colors looking about right in photographs, whether by AWB or otherwise. Dicklyon 23:04, 18 May 2007 (UTC)


 * And when a camera "adapts" to the illuminant, it's called color|white|gray balance, to distinguish it from the human visual phenomenon of chromatic adaptation. That's how it has been explained to me. Lovibond 23:33, 18 May 2007 (UTC)


 * I don't think it's that simple, but I'll look for some refs. Dicklyon 23:43, 18 May 2007 (UTC)


 * Nothing ever is that simple, is it? Lovibond 23:30, 19 May 2007 (UTC)


 * This difficulty in determining the precise transformation implied by each term is the reason I originally suggested merging "white point" into this article. It would be good in my opinion to have a summary somewhere, explaining what different types of mathematical adaptations can be made to remove color casts, and to explain any ambiguities of the terms involved.  I'm a complete novice at this stuff, so I can't help too much to describe what is meant (unless there are some particularly relevant papers I should try to read and help summarize; I can get most recent papers online through school), but I've found terms like "white balance," "color balance," "color temperature," "illuminant," paired with words like "transformation" and "adaptation" used with a whole range of intended meanings, and it is not always clear exactly what an author means by a particular such phrase.  I wonder if there is some source which clarifies the possible usages of these terms, and provides more precise terminology, so that we can avoid any confusion. --jacobolus (t) 06:55, 16 May 2007 (UTC)


 * I agree, we need to find some sources and see how they use the terms, and quote and reference them. I have a bunch of books to consult on this.  I'm not a novice at it, but I always go back to the books when these kinds of questions need to be resolved. Dicklyon 23:04, 18 May 2007 (UTC)


 * I have to side with the folks that said that the article is about what I could call "white balancing". Color balance is a broader concept... take a look at the original article that's at the top or just search Google for "color balance". For example, when a photographer says he wants to change the color balance, that might entail adding this color and subtracting that. Photoshop even has a filter called Color Balance, and it allows you to do just that. White balancing is a specific type of color balance operation; one you do to make neutral objects appear neutral.--Adoniscik (talk) 07:03, 21 January 2008 (UTC)


 * Well, speaking as an expert on white balance, including automatic white balance, and chromatic adaptation, I'd say that the notion that it's about making neutral objects appear neutral might be true, but it's not the same as rendering them as objectively neutral, which is why Photoshop and other apps try to make it easy to exercise manual control over the balance. But there's no real distinction in what the operation is.  In general, you pick a space to adjust the gains of the primaries in, whether you call it white balance or color balance.  If we make this article "white balance", I'm having a hard time imagining what the "color balance" article would be.  The alternative approach is to just make sure we say all that needs to be said about color balance, and then, if it's too much, we can consider a split. Dicklyon (talk) 08:10, 21 January 2008 (UTC)

how does this fit in with color correction ??
The article about color correction talks about using colored filters over the lens, etc. to change color balance. I wonder what the proper place is for discussing the use of techniques like filters, changing color exposure times on various colors in making (pre-digital) photographic prints, and other types of color corrections is. There's also some discussion of how color balancing is done in motion pictures in the color grading article. I wonder if any of these articles should be put as sub-sections (w/ main article link) of each-other. Perhaps color correction should be the top level such article, and should have sections describing the meaning of color balance, color temperature, white point, etc? --jacobolus (t) 07:32, 22 May 2007 (UTC)

Progress thus far (23 May 2007)
Wow, folks -- great discussion, and how the article itself has grown. Lovibond 20:36, 23 May 2007 (UTC)


 * The two of you have indeed done a nice job so far. I'm happy to sit and cheer from the sidelines. :) --jacobolus (t) 00:31, 24 May 2007 (UTC)

Choosing the neutral point
This article is helpful, but I don’t learn from this how the camera chooses the neutral point to use to (automatically) calibrate the white balance. If there is a black street, a green tree, and a white cloud, in the field of view how does it decide what pixel(s) to peg to neutral? Thanks! --Lbeaumont 23:17, 6 November 2007 (UTC)


 * Several AWB methods are mentioned, and several have references; but you're right it doesn't say, because AWB should be a whole article on its own. There's no method that always works well. Dicklyon 07:05, 7 November 2007 (UTC)


 * I came here to see what it said about AWB, especially related to film and scanning. As far as I know, back to the early color negative printing days, there is a system to color balance based on the overall color of an image.  One way was to put a diffuser over the lens, and then read off the average color with a light meter and red, green, and blue filters. As well as I know, this was commonly done for non-professional color negative printing.  It sometimes fails when the image has large blocks of one color, though. I wanted to see what this article would say, both from the color negative printing process and AWB in digital cameras. Gah4 (talk) 17:48, 29 October 2018 (UTC)

Color balance using other than RGB
I wonder whether “adjustment of the relative amounts of red, green, and blue primary colors in an image such that neutral colors are reproduced correctly” is really always the proper definition here. Specifically, I would call the “tint” and “temperature” sliders in Adobe Camera RAW “color balance” tools, but they don’t really change relative amounts of RGB except indirectly. (As far as I can tell, they basically shift the image in a* and b* dimensions.) So I suppose the question is, do chromaticity corrections in non-RGB models count as “color balancing”? —jacobolus (t) 14:25, 30 April 2008 (UTC)


 * How do you even know what those controls do? If they do change RGB balance, even if indirectly, then the definition seems OK.  But I'm open to a more general definition if you can find one.  Dicklyon (talk) 15:47, 30 April 2008 (UTC)

Don't they indicate what chromaticity to perform chromatic adaptation to? The temperature is self explanatory; the tint is the deviation from the Planckian locus, I would imagine. --Adoniscik(t, c) 15:53, 30 April 2008 (UTC)


 * Yes, they have parameterized color balance by a color temp and a tint. Dicklyon (talk) 16:12, 30 April 2008 (UTC)

Going further, it seems that "tint" is $$\Delta_{uv} \times 3000$$. --Adoniscik(t, c) 16:25, 30 April 2008 (UTC)
 * You mean Photoshop's arbitrary representation of tint as one chromaticity dimension, according to the guy who should know. Fine. Dicklyon (talk) 16:39, 30 April 2008 (UTC)

From the OP, I thought we were specifically discussing Adobe Camera RAW? --Adoniscik(t, c) 16:46, 30 April 2008 (UTC)


 * Right, ACR. And he's talking about "chromatic adaptation", which is one good approach to color balance, but certainly not the only one. Dicklyon (talk) 20:58, 30 April 2008 (UTC)

Okay, I guess the reason that I am a bit unsatisfied with the definition is that such controls aren’t just shifting relative amounts of R, G, and B, but are (as far as I can tell) changing chromaticity while leaving luminance alone. Just shifting relative amounts of R, G, and B doesn't do that. (Thanks for the link Adoniscik.) —jacobolus (t) 01:15, 1 May 2008 (UTC)


 * It depends on what you understand by "color balance". As a Photoshop user, I would lump all the color correction tools under "color balance" and the ones you mentioned under "white balance". It seems everybody has their own definition, and it's easy to find references backing each position. I consider "white balance" a special case of "color balance". I think the confusion partly derives from the usage of balance in casual phrases like "the balance of color in this picture is wrong", which then gets contracted to "the color balance of the picture is wrong". The term "White balance" does not lend itself to such usage, and therefore has a narrower meaning. --Adoniscik(t, c) 03:59, 1 May 2008 (UTC)


 * Interesting. Show me some of those refs.  I would not have thought that's a difference between white balance and color balance, or that a bunch of different tools would be lumped under color balance.  I don't understand what it is that you don't like about the book you linked. Dicklyon (talk) 04:04, 1 May 2008 (UTC)

I did not say I did not like it. I more or less quoted a sentence that uses "color balance" in the sense "relative amounts of R, G, B". If you ask a visual artist not versed in imaging terminology they too will likely say the same thing. We are trying to appropriate a term that is already in popular use:


 * The color balance in an image refers to the overall proportion of various colors in an image


 * Learning how to adjust the white balance is the best initial method of achieving the desired color balance of the final image (note the deliberate distinction)


 * In its broadest meaning, color balance is an aesthetic term. (Munsell)


 * The term color balance refers to the relationship between the colors in a print or transparency and those in the original scene...


 * Color Temperature Color Temperature adjusts the picture's color balance along an axis that begins with red and ends with blue.

I'll elide the countless references to "Color Balance" filters in various imaging applications, including Photoshop, that are used to adjust the relative amounts of R,G,B. --Adoniscik(t, c) 04:44, 1 May 2008 (UTC)


 * Sure. I'm not denying that “color balance” tools are used to adjust relative amounts of R, G, B.  I’m just wondering a) how general the definition of “color balance” is, b) if there’s a general term which encompasses both types of adjustments, and c) what the proper term is for color corrections more sophisticated than simple R/G/B twiddling.  Then this article should be placed IMO at whatever the most general term is, or if usage varies or is ambiguous, that should be explained in the article text.  &c. —jacobolus (t) 06:07, 1 May 2008 (UTC)


 * Well, it's certainly the case that usage is variable and ambiguous, such that even with lots of sources it's going to be tricky to explain. As to the one that Adoniscik found with both "color balance" and "white balance" in one sentence, I can't find much clue as to what they think those mean, or what the distinction is that they think they're drawing between them; the whole paragraph is content free. Dicklyon (talk) 06:15, 1 May 2008 (UTC)

Not saturation ! Definition needs improvement.
Delete the box to the right?

Hmmm ... looks like 'sun vs shade' to me, not 'accurate vs vivid' !

Although, that is understandable in the context of a poor definition
 * color balance is the global adjustment of the intensities of the colors

which needs this clarification
 * (typically red, green, and blue primary colors).

'Intensities' is the wrong word - 'levels' or 'quantities' or 'amounts' would be better.

Or do artists, or one camera manufacturer use a wider definition ? --195.137.93.171 (talk) 11:18, 6 September 2011 (UTC)


 * PS I was unwittingly redirected here from White Balance - so you see whence I am coming !--195.137.93.171 (talk) 11:21, 6 September 2011 (UTC)


 * OK - I've read around quite a bit. Problem is semantics.
 * 1) The article seems to have totally lost the artistic/subjective/aesthetic definition - essentially 'colour harmony'
 * "In artworks in general, the color palette can be selected by the artist to convey a mood or an impression."
 * eg Picasso's Blue + Rose Periods, choice of (complementary or contrasting, but not clashing) color palette in web-page or other graphic design.
 * I guess one could make a distinction between white balance ...
 * 2) "grey surfaces should look grey"
 * ... and colour balance
 * 3) "things should be represented as they are in life"
 * eg don't correct for lighting colour - so things don't appear as if white-lit - things lit red look red
 * 4) I suppose also one could 'balance' so that summing the colours in a picture gives grey ! "chromatic adaptation" ?
 * All pretty subjective, but luckily the brain is good at interpreting colours.
 * Anyway, saturation is the last of the colour dimensions that I would consider as 'balance' !
 * --195.137.93.171 (talk) 14:34, 6 September 2011 (UTC)

What does the icon represent ?
I was wondering what the icon on camera menus represented, and found this. Or the Crossfader on a Disc jockey Mixer - used to fade out the end of one song, and seamlessly fade-in the beginning of the next.
 * The sliders in stereo amplifiers that you use to (manually) set the balance between the left and right audio channels?

Anyone have a better idea, or even a 'reliable source' ?

It may represent something obsolete, like the Floppy disk icon for 'Save' !

--195.137.93.171 (talk) 10:56, 4 July 2012 (UTC)

Assessment comment
Substituted at 12:04, 29 April 2016 (UTC)

Example of extreme white balance adjustment?
Not sure whether this would add much value or not, also unsure as to policy on embedding animations in to wiki pages, so I though I'd post this in the talk section and see what others think rather than add it to the main article. Intent was to provide an example of a really dramatic adjustment, with the initial features being almost invisible and also showing how noise and other visual issues can be increased or hidden. Thoughts? Prosthetic Head (talk) 13:17, 29 April 2016 (UTC)

pre and post
It seems to me that the article could mention, or just follow along, with the distinction between pre- and post- color balance. In the film days, this would be filters over the lens when taking the picture, or filters when printing from color negatives. (With color positives, one is expected to get it right in the first place.). Because of non-linearities in film, it is much better to get right in the first place, and so negative films balanced appropriately were created. But because of differences in printing paper and printing light sources, one still must balance at printing time. One part of the article mentions filters over the lens, but not what can be done at printing time. In the digital case, again one can change the balance later, but it is best to get it right (or close enough) in the first place. One can then make fine distinctions later. Gah4 (talk) 17:56, 29 October 2018 (UTC)