What is the Definition of Color According to the CIE?

[D H]

The second definition is the correct one.

I disagree. Neither is the correct definition. I'll highlight the part of FactChecker's post with which I disagree.

Lets compare two ways we could define a color:

The single frequency number: A single light frequency has a given effect on our eyes, specific physical properties when filters and lenses are used, and a linear progression from infrared to ultraviolet as the numbers increase. All these properties are independent of intensity.

The table of frequency mixtures: We could define a color as a huge table of mixtures of multiple frequencies, mixed at specific proportions that depend on total intensity. They must all have the same effect on a "standard" human eye. To a non-standard eye, even members of the same table would appear to be different colors. They would have a very complicated mix of behaviors when filters and lenses are used. There is no practical way to order the mixtures in a progression from infrared to ultraviolet.

There most certainly is a practical way. There are many practical ways. One is the CIE color space. This has been mentioned multiple times in this thread. Here's a slice of the 1931 CIE color space:

The curved boundary represents the spectral colors. Notice how narrow yellow is, and if you look at the frequencies, how narrow blue/violet and red are. On the other hand, the shades of green occupy a good portion of the CIE color space. Our eyes are very good at detecting shades of green. They're not so good at detecting shades of blue/violet, or red. Finally, that flat line at the bottom: That's the line of purples. There is no such thing as "purple" light. That does not mean that there is no such color as purple.


Here's another: What color is the center block?

If you are a normally sighted person, that center block is "yellow". Yet there is barely any "yellow" light coming from your computer screen. It is instead a combination of red and green light.

The ways in which your computer monitor, your color TV, and your color photographs reproduce colors have been very carefully calibrated to reproduce color as perceived by a normally sighted person.

 

[sophiecentaur]

What is really "green"?

Hmm. I would have said that we often misinterpret what we see, on account of the colour. Take the chameleon as an example. Or even colour TV? But I don't think we're arguing over very much at this stage.
Edit: this in answer to the objection to colour being an illusion.

A point about the CIE chart is that equal distances at different places do not represent equal steps in perception steps. There is a modified chart but I'm on a train and I can't locate it.

 

[dauto]

Hmm. I would have said that we often misinterpret what we see, on account of the colour. Take the chameleon as an example. Or even colour TV? But I don't think we're arguing over very much at this stage.

No, I think we are arguing about the very definition of the word color. D H 's got it right. color is defined in a 2-D space. any combination of visible light can be mapped to that 2-D space. If you want to understand color you have to study that 2-D space and the mapping between light spectrum and the 2-D space. What doesn't make sense is to declare color as identical to wavelength and declare the perception of color as an illusion. That's an easy way out that provides no true understanding about what color really is.

 

[dauto]

I disagree. Neither is the correct definition. I'll highlight the part of FactChecker's post with which I disagree.

I agree. his second definition is better than the first, but still imperfect. Baby steps...

 

[sophiecentaur]

No, I think we are arguing about the very definition of the word color. D H 's got it right. color is defined in a 2-D space. any combination of visible light can be mapped to that 2-D space. If you want to understand color you have to study that 2-D space and the mapping between light spectrum and the 2-D space. What doesn't make sense is to declare color as identical to wavelength and declare the perception of color as an illusion. That's an easy way out that provides no true understanding about what color really is.

But the CIE chart only shows the outputs of three filters. The mapping is only one way. Two (three) numbers cannot describe the detailed received spectrum. I think the word for this could be degenerate. CIE doesn't do any better than your brain, in any meaningful way.

 

[dauto]

But the CIE chart only shows the outputs of three filters. The mapping is only one way. Two (three) numbers cannot describe the detailed received spectrum. I think the word for this could be degenerate. CIE doesn't do any better than your brain, in any meaningful way.

That's exactly right. The color map is degenerate. It is what it is. That's why we do not rely on color to describe a spectrum. It ain't good enough.

 

[Pythagorean]

Hmm. I would have said that we often misinterpret what we see, on account of the colour. Take the chameleon as an example. Or even colour TV? But I don't think we're arguing over very much at this stage.

At some point, we have to have consistent definitions in order to claim there's an illusion: if there's no reality to compare the illusion to then there's no basis for calling it an illusion. So we define color as DH has been careful to use it: the agreed perception of a "normally sighted" person. And we like this definition, especially since we people aren't "normally sighted" we can pinpoint, physiologically, why.

Now, if you look at DH's post too carefully, you could say "Well, DH differentiated between light that is really red and green, and the perception of them together as yellow". So, if that's what you're studying, you could call the yellow perception an illusion, but then you're saying that the "real colors" are red and green. But those labels "red" and "green" are defined, also, by a combination of collective perceptions and wavelength and are taken to not be illusions in this case (and, in fact, play the role of the reality to which we compare the illusion to!).

So, ultimately, it'a matter of context: it depends on what aspect of perception you want to measure with your illusion.

 

[sophiecentaur]

I think our views on the meaning of the wired 'illusion' are not identical. A moving colour TV picture is a total illusion, spatially, temporally and colourimetrically. All three sensations we get from the screen are not really what is in the original scene. If you don't spot a banana, placed on a yellow carpet, that's because we have the illusion that the spectra of both are the same - i.e. in our limited way we cannot distinguish them. A small cow in front of a larger cow can look the same size. Without other visual clues, we can't resolve that illusion.
I cannot see there is any difference between the two illusionary experiences.
At least we can agree that 'colour = wavelength' is total nonsense, which is our main point here.

 

[Pythagorean]

My meaning for optical illusion is derived from vision sciences, where illusions serve a specific purpose, I believe you are using it more casually. Anyhow, yes, we do agree that wavelength = color is nonsense, though it's important to note that wavelength helps determine what colors are what (though our perception plays a major role there, too, e.g., in defining a dividing line between what's red light and what's orange light, and different cultures place that line differently). I wouldn't go as far as to say color is an illusion. Color is a sensory experience and we can create illusions with color (sometimes those illusions are themselves color, but not all sensory experience of color are an illusion).

But if we call all sensory experiences illusions, then we lose the whole purpose of illusions in the cognitive sciences. Illusions are more about how sensory experiences are interpreted. I would agree that movies are an illusion. Thinking the banana is not there is also an illusion. Your examples seem to be illusions. But that is not the same as saying color is an illusion.

 

[sophiecentaur]

I can go along with "sensory experience" OK. Of course, all our inputs are the same and they are all subject to misinterpretation and the mapping is always degenerate.
Colour seems to be an area of the senses that is very poorly appreciated yet people seem more prepared to express their own primitive model confidently than they do in other areas. I guess it's the first phenomenon that we are taught about in any quantitative way. That could account for the over-familiarity.

 

[FactChecker]

Human sensory perception is not a good basis for a definition. We do not define speed by how tired we get running that fast. We do not define weight by how much it hurts to lift something. We do not define hardness by how it feels to get hit by something. Those may have motivated the concepts for primitive man, but they are not satisfactory for the modern world. All of those have better definitions that are widely accepted. Color is the same.

 

[FactChecker]

I disagree. Neither is the correct definition. I'll highlight the part of FactChecker's post with which I disagree.
There most certainly is a practical way. There are many practical ways. One is the CIE color space. This has been mentioned multiple times in this thread. Here's a slice of the 1931 CIE color space:

The curved boundary represents the spectral colors. Notice how narrow yellow is, and if you look at the frequencies, how narrow blue/violet and red are. On the other hand, the shades of green occupy a good portion of the CIE color space. Our eyes are very good at detecting shades of green. They're not so good at detecting shades of blue/violet, or red. Finally, that flat line at the bottom: That's the line of purples. There is no such thing as "purple" light. That does not mean that there is no such color as purple.Here's another: What color is the center block?

If you are a normally sighted person, that center block is "yellow". Yet there is barely any "yellow" light coming from your computer screen. It is instead a combination of red and green light.

The ways in which your computer monitor, your color TV, and your color photographs reproduce colors have been very carefully calibrated to reproduce color as perceived by a normally sighted person.

Your diagram makes my point. Every color on your chart appears in a single line ordered by frequency. The two dimensions are not needed. They just confuse things. Furthermore, experiments have shown that there are single-frequency colors that cannot be exactly duplicated with any mixture of primary colors. Experimental subjects could tell the difference no matter how the mixture of primary colors was adjusted. Furthermore, there are other color spaces defined that have a great deal more colors than your chart.

 

[AlephZero]

Your diagram makes my point. Every color on your chart appears in a single line ordered by frequency.

Sorry, that is wrong.

The straight line along the bottom of the chart (including purple) is NOT ordered by frequency, or by anything else.

 

[FactChecker]

Sorry, that is wrong.

The straight line along the bottom of the chart (including purple) is NOT ordered by frequency, or by anything else.

I stand corrected and will grant your point that the purples are not on the single-frequency curved exterior line. And that makes sense. Purple is a color where the reds will stimulate the low frequency eye sensors and the blues will stimulate the high frequency eye sensors, but the middle frequency eye sensors are not stimulated. No single frequency color can do that.

I believe that the OP was asking if the combination of distinct colors, yellow and blue, could really generate the frequency of the single-frequency color, green. The answer is no. The single-frequency colors are the extreme points of the convex set of colors. They can be combined to make all other colors but other colors can not be combined to make a single-frequency color. The chart shows that clearly. So there is still a strong case for defining the single-frequency colors first and defining all other colors as combinations of the single-frequency colors

 

[bahamagreen]

Perception is phenomenological and very hard to unravel analytically. Things are not what they seem and virtually every assumption that has been posed about how it works has fallen miserably short of squaring with what little is actually known... already known to be quite strange and counter intuitive. Color perception is just the tip of the iceberg; all the perceptual modalities have amazingly peculiar aspects, especially when their "components" are examined.

For example, if you play catch with a yellow tennis ball, you see the yellow ball moving as a whole and coherent perception - good enough to throw to the other and catch when thrown to you, but the actual locations in the brain that process and extract and present those three features (that it is a ball, that it is yellow, and that it is moving) are spatially separate and different parts of the brain... shape, color, and motion seem to be determined independently (and invisibly or "non perceptually"), yet these "components" become fairly perfectly integrated somehow so that you perceive a coherent moving yellow ball. There appears to be no one place in the brain where all three features of the moving yellow ball "coincide", yet you see what you see.

There is much going on here that is barely understood at all.

 

[OmCheeto]

Well ok then. I always though Violet and Purple were just different shades of a similar color.

I didn't know that Purple didn't really exist, except in our heads.

Please ignore all of my previous posts.

 

[Pythagorean]

All of those have better definitions that are widely accepted. Color is the same.

Well, if you haven't gotten the memo by now, I don't think another round is going to help much...

 

[bobie]

I believe that the OP was asking if the combination of distinct colors, yellow and blue, could really generate the frequency of the single-frequency color, green. The answer is no
The single-frequency colors are the extreme points of the convex set of colors. They can be combined to make all other colors but other colors can not be combined to make a single-frequency color. The chart shows that clearly. So there is still a strong case for defining the single-frequency colors first and defining all other colors as combinations of the single-frequency colors

I mentioned green, but I then realized I should have asked about yellow. I knew that two EMR frequencies do not interfere or add up, but I naively thought that mixing yellow and blue paint would produce a chemical compound emitting a single frequency, which was a plausible guess. But you said that two frequencies survive side by side and are processed in the mind.

Nobody has ever commented my attempt to define 'objectively' colour-frequency , drawing a parallel with 'note-frequency, on which there is no strong debate'.
I pointed out that sound-light are both frequencies that can be defined objectively, with a definite number of Hz.
What we do is : we arbitrarily set boundaries , group frequencies and define single/basic notes-colours : A,B..., red, blue...
Then we combine basic elements and we get new effects in our mind that acquire an individuality, a particular 'taste' : like a diminished fifth , ACG etc in sound or purple in light.

The only difference is that whilst with sounds a combination of notes always gives a new/original 'sensation',
with colours two frequencies, (in a few instances, as red (,blue etc..) is never a false interpretation), are interpreted as a single frequency: "yellow".
The scheme seemed flawless to me, nobody showed its shorcomings. Why can't it be as simple as that?

 

[sophiecentaur]

I mentioned green, but I then realized I should have asked about yellow. I knew that two EMR frequencies do not interfere or add up, but I naively thought that mixing yellow and blue paint would produce a chemical compound emitting a single frequency, which was a plausible guess. But you said that two frequencies survive side by side and are processed in the mind.

Nobody has ever commented my attempt to define 'objectively' colour-frequency , drawing a parallel with 'note-frequency, on which there is no strong debate'.
I pointed out that sound-light are both frequencies that can be defined objectively, with a definite number of Hz.
What we do is : we arbitrarily set boundaries , group frequencies and define single/basic notes-colours : A,B..., red, blue...
Then we combine basic elements and we get new effects in our mind that acquire an individuality, a particular 'taste' : like a diminished fifth , ACG etc in sound or purple in light.

The only difference is that whilst with sounds a combination of notes always gives a new/original 'sensation',
with colours two frequencies, (in a few instances, as red (,blue etc..) is never a false interpretation), are interpreted as a single frequency: "yellow".
The scheme seemed flawless to me, nobody showed its shorcomings. Why can't it be as simple as that?

This is amazing. After more than 80 posts you are still claiming that we see "frequencies. Two spectral colours lie on that curved part of the CIE chart. A weighted sum of two colours (wherever they are on the CIE chart) will produce a colour will lie on a straight line between those colour points. This colour cannot actually lie on the arc of spectral colours (frequencies). So even that very simple example does not produce perfect mimicry of a third spectral colour. The nature of our vision is that we can mimic, quite well, the vast majority of colours, using combinations of three (RGB) primaries. I will reiterate my point that the primaries that are commonly used are not spectral colours (they don't lie on the curve) so there is absolutely no chance that the yellow they can produce will be spectral frequency which corresponds to what we call 'yellow'. I spent quite some time with colourimetry in the pass and I would say that I cannot remember any instance when the word 'frequency' or 'wavelength' was used in any of the studies, except to describe that curved perimeter - and that is outside the gamut of any synthesis system.

I have come to the conclusion that the words colour / frequency must have become a shibboleth pair to separate people who do and people who don't have a history in colourimetry. Can it all be really that difficult to take on board?

 

[Pythagorean]

Nobody has ever commented my attempt to define 'objectively' colour-frequency , drawing a parallel with 'note-frequency, on which there is no strong debate'.

Sure they have. DH presented the very framework for objectivity itself, the CIE color space. And I mentioned that similar perceptual abstractions are present in human sound processing.

One of the differences between objective pressure waves measured in Hertz and the notes as humans hear them leads to an error called the syntonic comma; that we perceive twice the frequency as the "same" note (an octave) is a perceptual artifact itself: that leads to something like a diminished fifth sounding dissonant, and a major or minor triad consonant. In fact, the nicest sounding notes (in relationship to each other) are whole integer ratios of each other (1:2, 2:3, 3:4, etc), which follows from the harmonic series.

The nature of the perceptions are rooted in harmonic theory and resonance, but in perceiving it, our brain makes many abstractions of the raw data. Also, realize this: you only have two sensors for sound and you process it as a single dimension. You have a spatial distribution of millions of cone cells that process the 3D landscape + color-coding that you experience.

Anyway, people already objectively define color. That's the whole point of the CIE Color Space. We also have Opponent-Process Theory that gives us a mechanism for how human's interpret colors and helps explain why some humans can't distinguish certain colors (color blindness) which allows us to design tests to detect when people are color blind.

 

[bahamagreen]

Let's be clear about any sense of physical frequency with respect to neural function and perception.

The fastest speed at which neural membranes can depolarize ("fire") and wait out their refractory period (recovery until possible to fire again) allows a top firing rate of about 1000 Hz. There are no neural signal rates faster than that, most are slower.

Physical frequencies of light are NOT being sensed whatsoever.

Retinal molecules (pigments) are "unbleached" using energy into a bent (cis) shape. They are cocked like a spring waiting to be released. When a photon is absorbed at the 11-cis-retinal chromophore (the "spring release trigger"), that 11-cis state changes shape back to all-trans. This is called "bleaching" of the pigment. Individually, that is not enough to initiate anything... it take about 6 photons being absorbed close together at about the same time for the retinal molecules involved, residing within in a disc-like wafer, these discs stacked in the cones, to kick up enough activity to initiate a depolarization. It is the specific protein bound to the chromophore that determines what frequencies will be absorbed. The signal from the cone only indicates, "some photons were absorbed"...

 

[bobie]

After more than 80 posts you are still claiming that we see "frequencies.

I said "interpret".

 

[FactChecker]

The CIE 1931 color space is mathematically calculated by combining single-color spectral colors. It is exactly the mathematical definition that I have said is preferable to defining colors by a sensory perception. But it does use weighting functions that represent the "standard observer". That is both a blessing and a curse. It makes it easier to understand how the color will be seen by the "standard observer" and it formally defines the "standard observer" but it makes it dependent on that definition, and therefore debatable. Other CIE versions use other weighting function definitions of the "standard observer". Defining colors by perception is like saying "I can't tell you what it is, but I know it when I see it".

The most saturated red, green, and blue are spectral colors. Experiments have shown that people can tell the (sometimes subtle) difference between the saturated spectral colors and colors made from mixed primaries. The saturated spectral colors are not included in common color spaces like Adobe RGB, CMyk, and sRGB.

P.S. We cannot see the CIE 1931 color space on computer monitors. Monitors can only show the subset of colors in its smaller color space. You have to purchase sheets to see the real thing.

 

[FactChecker]

But the CIE chart only shows the outputs of three filters. The mapping is only one way. Two (three) numbers cannot describe the detailed received spectrum. I think the word for this could be degenerate. CIE doesn't do any better than your brain, in any meaningful way.

The CIE 1931 color space includes all the visible spectral colors. The color spaces build from the three primary colors (Adobe RGB, sRGB, etc) are subsets. Even the three primary colors used in those smaller color spaces are not the fully saturated colors in CIE 1931.

 

[sophiecentaur]

What is really "green"?

The CIE 1931 color space includes all the visible spectral colors. The color spaces build from the three primary colors (Adobe RGB, sRGB, etc) are subsets. Even the three primary colors used in those smaller color spaces are not the fully saturated colors in CIE 1931.

I'm not sure whether you are disagreeing or agreeing here.

 

[dauto]

The CIE 1931 color space is mathematically calculated by combining single-color spectral colors. It is exactly the mathematical definition that I have said is preferable to defining colors by a sensory perception. But it does use weighting functions that represent the "standard observer". That is both a blessing and a curse. It makes it easier to understand how the color will be seen by the "standard observer" and it formally defines the "standard observer" but it makes it dependent on that definition, and therefore debatable. Other CIE versions use other weighting function definitions of the "standard observer". Defining colors by perception is like saying "I can't tell you what it is, but I know it when I see it".

Yes that's exactly right. color is defined by perception which makes it a hard to pin point concept. It is what it is. It is not good enough for more analytical purposes and that's why physicists rely on spectrograph, not just color. No matter how you slice it, color is entangled with the human eye and brain perception process. Pretending it to be otherwise doesn't help anything.

 

[sophiecentaur]

Human sensory perception is not a good basis for a definition. We do not define speed by how tired we get running that fast. We do not define weight by how much it hurts to lift something. We do not define hardness by how it feels to get hit by something. Those may have motivated the concepts for primitive man, but they are not satisfactory for the modern world. All of those have better definitions that are widely accepted. Color is the same.

On the planet zog, they could be using exactly the same data for working out speeds, masses etc. All they would need to have done is what meteorologists do (or aim at doing) on Earth - namely to base units on things like light speed and carbon atoms. There is no way that we could expect the Zogians to have an identical CIE chart because it is entirely the result of phychophysical measurements using Humans with human eyes. Each sighted animal on Earth would have a different version of a CIE chart, too - if you could invent suitable tests. That makes Colour fundamentally different from your other examples of quantities. Colour is not the "same".

Edit: the CIE chart was not "mathematically calculated" from measurements. It was derived from many subjective tests.

 

[sophiecentaur]

I said "interpret".

Fair enough - but it isn't just "frequencies" when we see a colour; it is the weighted sum of the components that counts - a suitable term would be 'spectrum'. That one word of yours suggests that you are still not getting an important point. (Several other members have commented on this.)

 

[sophiecentaur]

@Factchecker

P.S. We cannot see the CIE 1931 color space on computer monitors. Monitors can only show the subset of colors in its smaller color space. You have to purchase sheets to see the real thing.

The "real thing" does not exist. You cannot produce spectral colours with pigments, although it is true that you can get more saturated colours with gels than with (standard) phosphors. The saturation is achieved at the expense of Luminance. The only "real thing" would have to consist of an array of numerous phosphors and a set of monochromatic sources along the curve.

The Pantone colours are 'standard' colours for dyes and they are accurately specified. However, they do not cover the whole CIE gamut. Any chart or card you could buy would need to be illuminated with the appropriate illuminant, of course, before you could rely on it.

 

[Pythagorean]

In pigments, I assume, a lot of the color comes from diffusive reflection and refraction throughout the material, leading to a lot more blending and chromatic dispersion as compared to a source like phosphors?

 

[sophiecentaur]

It comes mainly from the fact that a dense medium has very broad band reflection/transmission characteristics. Overlapping two passbands by a small amount will exclude most of the incident light but what gets through can be narrow band but low luminance. Under exposed colour prints tend to have saturated colours due to the non linearity of the sensitivity curves.
A transmission filter is much much better than a reflective surface due to the reasons you give although viewing a single coloured card under totally white light would be immune to corruption. Gloss finish photos tend to have 'deeper' colours under the best (non specular) lighting.

 

[FactChecker]

The various CIE color spaces introduce a variety of human "standard observer" weighting functions. But that is not to define colors. It is to define the perception of a human "standard observer". What about other animals like a bird or a bee? The CIE is meaningless for them. Does that mean color does not exist for them and only exists for a human? That is nonsense. The spectral density of light is universally accepted in physics and it says nothing about a "standard observer". That is the legitimate definition of colors. Colors are independent of who or what is observing it, be it a man, a bird, or a photocell.

 

[Dale]

This thread is long since started going in circles. Color is perceptual, frequency is physical. End of story.

Human sensory perception is not a good basis for a definition.

It is a perfect basis for terms pertaining to sensory perception. Human sensory perception exists and we need to define words used to talk about it. So obviously some definitions must be based on human sensory perception.

Color is a perceptual concept, not a physical one. On PF we don't get to redefine standard terms simply because we don't like the standard definitions. Regardless of your preferences the CIE is the standard-setting organization and their definition of color is authoritative.

Thread closed.