Talk:Color constancy

Demonstration figures
The demonstration figures need work. The first is a lightness illusion -- not color constancy The second is unclear -- for color constancy, you need to have a different RGB color in the image appearing to be the same perceptual color to the viewer. In this case, the RGB values are different, but the colors also look different.

From the article:
 * early morning daylight, whose main component is bluish, and also in the late afternoon light, when the main component is reddish.

You sure about that? The color of sunlight on Earth's surface depends on the amount of atmosphere it's passing through, ie the angle of elevation of the Sun above the horizon. Early morning and late afternoon should look the same.


 * Perhaps the author of that passage doesn't get up early enough to be aware of the sunrise phenomenon. :) Or else is referring to pre-dawn light, which will lack direct sunlight and be deflected (bluish) light from the atmosphere... same as dusk. --Brion

"The effect was discovered in 1971 by Edwin Land."

(The inventor of the Polaroid?)

I seriously wonder if he discovered it. He may have theorized it. But every photographer knew it before. If you take photo in artificial light with a daylight slide film the picture is very orange.

Ericd 01:17 Apr 7, 2003 (UTC)

At second thought this has a lot of connection with photography aren't retinex algorithms an automatic white balance algorithms ? Ericd 01:22 Apr 7, 2003 (UTC)

Is the caption of the figure with the guy holding the cards accurate? Isn't the whole point of the article that you can add a shade (of pink, say) to an image and have it look the same even though the actual intensity of the pinks might be different? Maybe I'm just confused about the caption? 131.243.77.221 (talk) 01:12, 18 November 2007 (UTC)


 * Indeed, as i noted on the image's talk page, all colours but the pinks are looking different (at least to me). I hope a more knowledgeable user will answer your question, since IMO it has some merit. -- Jokes Free4Me (talk) 18:26, 30 June 2008 (UTC)
 * Funny: to me all the cards look the same. Red cast on #2 adds an illusion of higher contrast (darker dark areas) and lesser detail, but the hues are the same. NVO (talk) 21:31, 17 July 2008 (UTC)


 * Dr Edwin H Land (the title was honorary) did not invent "the Polaroid". He invented Polaroid sheet polarizing material, and the first simple system of instant photography.
 * His well-remembered (and controversial) article on color vision appeared in the May, 1959 issue of Scientific American. I don't know where the 1971 date comes from. Land admitted that others had made similar observations, but they had been overlooked or forgotten.
 * The kind of "color constancy" the theory describes is not the same as constancy under varying illumination. It's rather the opposite -- the same stimulus produces different colors, which flies in the face of common sense.
 * The retinex theory includes the fact that a wide range of colors can be produced from a single wavelength of light. This shows that "something" is going on in the eye and/or brain beyond a simple RGB analysis. (If Wikipedia has an article covering this, it's not easily locatable.)
 * Dr Land was attacked by scientists who couldn't swallow the implications of the retinex theory. Some of these attacks were personal, as Land had not completed his college studies. This article's casual dismissal of the "utility" of the retinex theory strikes me as yet another attack. WilliamSommerwerck (talk) 11:09, 7 April 2012 (UTC)


 * Indeed, I learned about the very same Land effect (that surrounding colors affect human color perception) in 1964, by that name, in a lecture I heard as a freshman in college. So I don't know the basis for the 1971 date, but it is considerably later than Land first discovered (or wrote about) this effect.Daqu (talk) 21:27, 25 June 2013 (UTC)

Is “Discounted” a special term?
I am having bum by the use of the word “discounted” in the phrase: “This illumination is then discounted” Is it a technical term? Can it be defined or “glossed” or better explained here? Would a word such as “normalized” be better? Would it be better to say that “the spectral characteristics of the illumination are estimated and this estimate is used to interpret the colors of the reflected light”? --Lbeaumont 13:55, 7 November 2007 (UTC)

"Discounted", a term that goes back to Helmholtz in the 19th century, is the most justifiable and model-independent term of which I am aware. The illumination is "discounted" if a scene looks the same under one illumination as under another. It is a matter of (I think misguided) modern dogma to color constancy as estimating the illuminant spectrum prior to compensating it. Surely retinex ratios don't do this, and neither do many other color-constancy algorithms including some that parameterize reflectance but not illuminant spectra. Michael H. Brill12.26.40.88 (talk) 16:35, 4 January 2008 (UTC)

Example of the man holding the colored cards
Well, the example seems a little ridiculous. Of course the cards seem to be of different colors in the two photos---at least one of the photos has been filtered. The colors are actually changed, not just apparently changed, and the viewer sees them as changed. So where is the illusion?CountMacula (talk) 11:45, 28 February 2015 (UTC)
 * Double checked, no you're wrong. -- CFCF  🍌 (email) 12:11, 28 February 2015 (UTC)
 * Oh, exactly how and what did you double check?CountMacula (talk) 15:27, 1 March 2015 (UTC)
 * I did a color-analysis in Gimp, and yupp they're the same. What it is you call a "filter" does not cover the pink color. -- CFCF  🍌 (email) 20:35, 2 March 2015 (UTC)
 * Ah, thank you. The caption use some clarification.CountMacula (talk) 11:45, 3 March 2015 (UTC)

article "White" -- more color science needed, including weighing in on current dispute about how to begin the lead (and ultimately other parts of the article)
Most of the editors of the article on the color "white" don't know what color constancy is, much less chromatic adaptation, and they don't even know there's a difference between "white light" vs a "white surface", or how these are related to the color "white". They see that article as being primarily about the cultural usage, figurative associations, and history of "white" (and "white" being no more fundamental than any other color like orange), rather than stating or explaining what white literally is as a color, which is well-understood in color science. Could any of you weigh in on the current dispute on the talk page talk:White there, which mainly is between just two editors about the lead and its first sentence? One of them wants the first paragraph to rely first on citing illustrative examples of white like snow, chalk, and milk, before even defining white in any objective sense, and based largely on one colloquial dictionary definition of the word "white". 129.33.253.142 (talk) 16:54, 4 October 2017 (UTC)

No such thing as a "gray wavelength"
The "Retinex theory" section speaks of "a photo with red and gray wavelengths." That doesn't make sense. There is no single "wavelength" of light that a person with normal vision would not perceive as a fully saturated color.