This is not a trick question.
Are there any opthamologists out there? I just want to know reliably how many colors the average human eye can see.
Thanks.
I am not the asked-for opthamologist but my copy of Photoghraphic Sensitometry - the study of tone reproduction (Hollis N. Todd & Richard D. Zakia 2nd edition 1974) states that “There are roughly 100 - 200 visibly different appearences within the visible spectrum.”
However there are three components to what we perceive as colour - hue, saturation and brightness. The above figure only refers to the first of these.
Oh and an explanation of those terms can be found here
Color is a wonderfully complex thing, as I’ve remarked before on this board.
In principle, there’s an infinite number of gradations of color – the visible spectrum is continuous, and our division into red, orange, yellow, etc. is artificial, just as there is a complete and continuoius range of aural frequencies. Your question seems to be about the number of distinguishable colors. I don’t know the answer, but ticker’s quote of 100-200 is probably right for most people. I’d be willing to bet that there are folks who can go beyong this, and that people can see other hues under the appropriate conditions. You can, for instance, see the red color of a laser that is nominally outside the range of human vision if it’s cranked up high enough (you look at the laser spot on a piece of paper, of course, not down the barrel of the laser).
You might want to consult David L. MacAdam’s books on Color Vision. He spent his life investigating just these questions. One of the things that makes this murky is that he found there are “color ellipses” within which lie all the color coordinates that people perceive as the same. (The color can be plotted out on a 2-D graph called the CIE Chromaticity Diagram). The thing is, I’ll bet the color ellipses overlap for some of those 100-200 gradations.
In a previous life I used to code image processing software. The 100-200 number for distinct hues is correct. But the human eye is much more sensitive to differences in intensity, so if you really want to know the total number of distinct colors we can perceive, the total can be very large indeed.
For example, many computer programs encode color using an 8 bit value for each of the red, green and blue primaries. (This is what you get when you set your monitor to 24 bit color.) This gives you about 16 million distinct colors. While this is adequate for most purposes, professional artists can easily distinguish between two 24 bit colors that differ only by a single bit. We discovered through a little experimentation that we needed at least 10 bits for each primary to erase all the distinctions. This puts an upper bound of ~1 billion on the number of distinct colors.
(Of course, we were going for overkill and I know we were wasting bits in the very bright and very dim parts of the range. I wouldn’t be surprised if the actual upper bound was closer to 100 - 200 million for a trained artist.)
A few notes:
-
RGB is a lousy set of primaries for data storage. It’s handy only because it maps easily onto the display hardware. It’s bad for storage because it’s not visually equal – one bit of data change produces different amounts of perceptual color change in different parts of the range. The numbers I give above were using a different set of primaries that were designed to be visually equal and more closely track the actual response curves of the cones in the eye.
-
Most CRT hardware can’t display more than 24 bit color anyway. The results we got were from scanning and printing high quality photographic transparencies using custom hardware.
-
We were calibrating our color to make maximum use of the dynamic range of photographic emulsion. Photographic emulsion and CRTs (indeed any color reproducing medium) can only reproduce a subset of the full visual range of human beings (think of a photo of a neon tube compared with the real thing). The range of a medium is limited to the maximum intensity and saturation of its primaries. So if you want to talk about the full range of colors the eye can see in the natural world you probably need to multiply my figure above by a factor of 2 or 3.
A lot.
Millions perhaps.
A new scanner I saw said it can do Trillions.
Who knows, if you change lighting to infrared, black light, etc, might be more than a couple million.
How would you count then anyway???
Is it even considered a color if the human eye can’t see it. The answer would then simply be all of them.
One thing I’m curious about. Color is supposedly a result of the mix of wavelengths in the visible spectrum reflected by an object. Conceivably, then, I could imagine a color corresponding to some continous function ranging between 0.0 and 1.0 over the frequencies in visible spectrum, representing the amount of light reflected at each frequency. Any of the color representation schemes, like RGB or HLS, can only possibly represent a very restricted part of this space. How do we know that this is adequate for color perception? How do I know there isn’t some funny distribution pattern I could create for the wavelengths in the visible spectrum that is not adequately represented as just a mix of red, green and blue, for instance?
I serviced TVs in the 70s.
I can tell you for certain that people do not see color the same. I used to ask the customer for their opinion on the facial tones when adjusting their set. It was always different from mine and always varied, sometimes dramatically.
I remember one call when the husband adjusted the color from behind the set. Both the red and blue guns were turned off and the whole screen was green. He actually thought he had a good picture.
In the 60s all males had to take selective service physicals. During one eye exam I remember them doing a color perception test. There was a number printed on a white card and each number was a different color. They were very faint and I missed one at first saying that the card was blank. I told the person giving the test he was wrong that there was not a number on that card so he would let me see it again. He showed me it again and with some embarassment noticed there was a pink 6 on the card.
I read somewhere that there was a theory that the ancient Greeks perceived hardly any colors and this is why Homer describes the sea as wine-dark. It is almost never that color. However, it could be they saw it that way because of no culture of color. The book also said that when you look at ancient Greek writings color is almost never mentioned at all, and if it is I think it said there were only 3 colors that were named. On the other hand, when you look closely with special equipment and Greek buildings and statues, you find that they painted them and that the metopes were blue and the triglyphs yellow, or vice versa, for instance. In addition, some societies have more names for colors than others but this may not mean they didn’t see colors as others did or do. Just because a person can see color doesn’t mean he will; generally people onlysee what they are told to see, as when the anthropologists sailed to the island and the natives asked where they came from. The anthropologists pointed to their ship, but the natives could not see anything. The anthros then explained what a ship was and how it worked and so forth and gradually the natives could see it. You can only perceive what you have a concept for (unless it hits you in the headd).
Concerning Cecils discussion of color.
What colors did the ancients use for paint in caves?
We can get into another round of how the ancients perceived color if we like - it’s a reasonably interesting topic. However, I would like comment on my question - how does our color perception relate to arbitrary distributions of frequencies along the visible spectrum, which could not be adequately described by a set of 3 parameters?
As I said – color theory is very complex.
As far as representing all seeable combinations using three primaries – they’ve done a LOT of testing on this, and they haven’t yet found a color they can’t represent on a Chromaticity chart. Theoretically, using three colors you can represent anything within the triangle of points they represent. (The three colors used for the CIE chart includes one “imaginary” color – it’s complicated.)
But it’s even worse than that. The human eye seems to respond in a tristimulus way, so you can get away with three-color technicolor and three phosphors on your TV set. But then there’s the Land effect, in which you can get full color reproduction from something that looks as if you only ought to see, say, various shades of pink. It was discovered by Edwin Land, founder of Polaroid, and is extremely weird. Look it up using a search engine.
On the other hand, tristimulus (making up everything using only three colors) is pretty weird itself. It means that you can have two different spectral distributions that can appear to be the same color to the human eye. This is NOT the way sound works – you can’t perceive sounds having two different spectral distributions as if they were the same sound.
If memory serves, the fellow that came up with the theory about the Greeks not perceiving color was Disraeli – the British Prime Minister of 100 years ago. His book pointed out the lack of color terms in the Iliad and the Odyssey. I agree with Cecil and numerous books that this is probably a quirk of language development, rather than a change in human perception over less than 3000 years. Julian Jaynes tried to use the same texts to prove that the Greeks weren’t conscious, but that’s another story.
Don’t the three primary colors coorespond to three different types of receptor cells in the eye, each with max sensitivity at that color? Since there are only three types of color receptors, three parameters are sufficient to describe any perceived color (not any color, but any perceived color).
Arjuna34
You’re right. The reason there are three primary colors (and not two or four) is because of the physiology of the human eye. If we had four different types of color receptors in our eyes instead of three we would need four primary colors as well.
Ultimately the color of an object is the continuous spectral distribution of its reflected light. But different spectral distributions can be peceived as the same color by the human eye. More receptors = more accurate color vision, i.e. a better ability to distinguish between unique spectral distributions.
Three receptors are the point of diminishing returns – there’s a big improvement in color vision when you go from two to three but much less when you go from three to four.
(BTW, the Red/Green/Blue primaries only roughly correspond to the response curves of the receptors. The receptors themselves respond to a wide spectral range and actually overlap with each other. So, for example, even “pure” green light will stimulate the red and blue cones to some extent.)
This an excellent discussion, and thank you for all the information. I will have to look into this more deeply.
I would like to tell you all why I asked. It was not just random curiosity. I don’t spend too much time in GQ, so many of you may not know me. I work in the technical field, and I always wondered why computer graphics cards went as high as 32-bit color when
A) I always thought many monitors were not capable of this
and
B) Many of my peers have said that “The human eye can’t perceive more than 16-bit color anyway”
It would seem, according to the information provided by Pochacco that the former may be correct and the latter may be erroneous. In any case, I thought it may help me get the answer I am looking for if I narrow the scope of my query.
So, without further ado:
When adjusting the display properties of your computer, you may notice that you can adjust the number of colors displayed by your system. The colors range from monochrome (black and white. I assume they call it monochrome because there is no black, just the absence of white) to True Color 32-bit. If 32-bit is hundreds of billions of colors, can we actually discern all these colors? For example, I have a very high-end video card in my computer, and when I set it to 32-bit color, I can see no difference whatsoever. I ask my friends, my girlfriend, my relatives and they also are unable to discern any difference. In any case, it would seem to me that we can’t see that many colors. Then why is it that video cards can do this?
Is it like cars that go 115 MPH, even when you will never conceivably go that fast? Is it like having 376 channels from your satellite service, even though you will never watch them all?
This is what made me ask this. I had been sitting around making “Hmmmmm.” noises, and finally decided to pose the question to the teeming millions.
Thanks for your consideration.
Yeah, thanks Pochacco and Arjuna34
Any idea what those response curves look like? Are they identical curves except for having maxima at different frequencies? Since some sort of chemical-based response is involved, I could imagine that this is not true. Is receptor response for a particular primary significantly narrower or broader?
I’ve been in the computer graphics field for the last seven years, so I think I’m qualified to cast some light (NPI) on the subject at hand. When talking about viewable colors on your monitor, we need to talk about two things: how your eyes see colors, and how your monitor/graphics card display colors.
The human eye has three kinds ofcolor-receptive cells, called the red, green, and blue cones. The red cones perceive wavelengths primarily between 470 and 680 nm, with the peak at about 580 nm. The green cones perceive wavelengths primarily between 440 and 630 nm, with the peak at about 545 nm, and the blue cones perceive wavelengths primarily between 380 and 510 nm, with the peak at about 440 nm.
Note that the red and green cones’ peaks do not correspond to the perceived colors of red and green, but, rather, yellow and greenish-yellow, respectively. The greenness/redness of a given color is determined by the difference between the green and red cones’ responses. Note also that the blue cone’s response at its peak is about 10% of the response of the green or red cone’s.
Now, let’s talk about the computer hardware. First of all, for every home-computer graphics card out there, there is no visual difference between 24-bit and 32-bit color. Both allocate 8 bits to each primary, red, green, and blue, per pixel. 32-bit color uses more memory, but for some graphics cards is drawn faster than 24-bit color.
Now, each primary value is intended to be a linear mapping of the color brightness. So, a value twice as large as another should be twice as bright as the other on the screen. But what actually is outputted to the CRT is the value for how strong the electron beam should be. Converting between the two, called gamma-correction, loses some bits of precision for the darker colors. Next, that binary value is converted to a voltage by a digital-to-analog converter, or DAC. Cheaper graphics boards will have only 6-bit DACs rather than 8-bit DACs, which can degrade the quality of the color reproduction even more.
To fully get 8 bits of precision after the gamma-correction step, you need about 11 bits of precision prior to the gamma-correction step. People sometimes fudge this to 10 bits, since the difference between 10 bits and 11 bits is often lost in the noise of the CRT, and since 3 10-bit values will fit in 4 bytes, but you need an extra bit to store 3 11-bit values.
Now, as far as seeing the difference between 16-bit and 32-bit color, you have to be careful setting up the test. If you have a very complex scene, you probably won’t notice the difference between the two. Simple scenes, with smooth color curves, on the other hand, will be very good at showing the differences. So, if you can set up a test scene, with a medium-gray background, and then a strip of gradiated color, about 100 pixels high, and at least 400 pixels wide, with the color on the left side being completely black, gradually fading to about 50% red on the right side, most people will be able to see the difference between 16-bit and 32-bit color.
I thought it was ophthalmologist instead of ophthamologist.
Mr. Sphygmomanometer