I'm always surprised to see barely high-school level science articles get on the front page.
An alternative title to this article could be "how spaceless numbers achieve defining space"? Like you just have a basis, and this defines space. Mind blown. Really, how do these magnets work?
It's not even an article I'd want to present to a high school class. An RGB pixel on your monitor is not emitting red, green and blue monochromatic (single wavelength) light. Same goes for the light reflected from a CYM pigment. Metamerism is at play, just like in the "real world". The author confuses the need to choose three primaries with a requirement those primaries be monochromatic. And since when have painters stuck to three pigments?
Painters don't stick to just three pigments, but all the pigments they use CAN be created by mixing three primaries in different proportions. I'm quite sure that's the whole basis for Thomas Young's colour mixing experiments performed in the 1800s, finding that every colour we see can be made with mixtures of three lights. Any more than that is redundant. Similarly, while an RGB pixel on your monitor may not emit all single-wavelength light, it doesn't negate the validity of the statement that three single wavelengths can be used to recreate all colours visible to the human eye. I'm quite sure that some RGB pixels are indeed single wavelength, though I would love to hear from those of you who know better about this because I don't know for sure...
>Painters don't stick to just three pigments, but all the pigments they use CAN be created by mixing three primaries in different proportions.
No. Paint is not light, and has physical properties that constrain what you can do with them. Some colors just don't mix well together, so you'll find recommendation on what exact type of pigment work best for specific mixes.
Yeah, that's true, paint is subtractive, as opposed to light. But isn't it still the case that all pigments can be achieved by mixing three primaries? Whereas in light mixing the primaries are considered to be red, green, blue, in pigment mixing they are considered to be cyan, magenta, yellow. Obviously please correct me if I'm wrong (no sarcasm implied), links would be useful.
No, I meant some pigments won't mix well because of other physical properties. The only example I can think of from the top of my head is in the Quiller, Color choices. It is mainly concerned with watercolors, and some colors mix uniformly (so blue+yellow -> green) and some look like an emulsion (so not green, but yellow with blue bubbles inside). Which can be interesting, because the mixing of colors is thus done by the eye instead of the paint, and is in fact one of the major painting effects used by impressionists.
No. No matter whether you're doing additive or subtractive color, there will be perceivable colors that you can't reach just by mixing a finite number of primaries. In fact, CMYK printing has a rather small gamut, with more colors that are clearly unreachable than RGB has.
Perceivable colors are a three-dimensional space, and colors you can make out of three primaries are a three-dimensional space, but they're different shapes. Mixing primary colors is a way of interpolating between them linearly, and the space of perceivable colors doesn't have convenient linear edges.
Here's an example. Your printer has cyan, magenta, and yellow ink (and also black for convenience). So it can print any color, right? It should be able to print bright lime green, something that looks like RGB #00ff00, right?
No, in fact, you can't print that color without "spot color" ink. When you mix cyan and yellow to make green, it will necessarily get darker. If you use enough cyan and yellow to get full-saturation green, it'll be too dark. If you use less ink, it won't be saturated enough.
Lime green is outside of the CMYK gamut. Similarly, the deep cyan you'd get by printing with lots of cyan ink is outside of the RGB gamut.
Not every school teaches this well, or at all. Not all people currently interested in this were listening to school lessons N years ago. Approachable explanations of everyday natural phenomena seem to be appropriate here to avoid narrowing your reading to startups and web tech.
>to avoid narrowing your reading to startups and web tech.
Where did I try to make this point? My issue is with the quality, not the subject. As pluteoid remarked, this is not a very good article, and fairly long-winded with regard to its content.
There has been many submissions of articles regarding color theory and its relation to psychology, design and physics, many of which are very interesting, and make you learn useful and applicable knowledge. For instance, you may learn how to write an imagemagick script to get the dichromat version of an image.
Right, but color (hue) shouldn't require a three-dimensional basis; it is just the frequency domain.
If each of the values of a three-dimensional coordinate pair encoded frequency, that triplet could carry information about a mixture of up to three colors. This would require sensors that are not "color blind".
I think that's a good summary of the article's point.
OP is incorrect; no tristimulus based (RGB or CMY) combination can reproduce the spectral locus without resorting to imaginary colour primaries. Imaginary primaries cannot be reproduced in reality, and as such, “all the colours we see” is “all the colours we may see from the printer.”
As a former neuroscientist who studied perceptual awareness, there's lots of little things wrong with this article.
1) The RGB palette used in monitors most certainly does not span the whole gamut of human vision. Not even the wide-gamut monitors used professionally can do that yet.
2) The diagram showing each cone type's peak spectral sensitivity is a bit misleading, too. The M- and L-cones are colored red and green, but if you notice carefully, their peaks lie in the green/yellow area, so characterizing them as red/green is misleading. And "each of our cones seeing a primary colour (blue, green, or red)" is flat-out wrong. Every cone fires in proportion to its overlap. The author clarifies this later, but it's best to tell people the caveats up front. Continually referring to "primary" colors is adding to the confusion here.
3) WhIle they did cite an article pointing out that ratio-computing cells exist in the retina, it does NOT follow that "it is the main source of information that the brain has to help it identify the wavelengths of light". The paper itself only talks about spectral separation in retinal cells pre-brain. I don't know where the author got the idea that ratios are the main source of info sent to the brain. Amplitude information is sent, as well.
I'm the author of the article, and I'm quite happy to see a response from a neuroscientist who studied this topic in the past. I wanted to ask you about point 3, because obviously I could have made a mistake. From what I understand (and what I have studied) retinal ganglion cells computing the S vs M and S vs M+L dimensions are the critical ones underlying coding of colour information along the blue-yellow and red-green axis. The additive axis, featuring amplitude information, from what I understand primarily concerns luminosity rather than colour discrimination. Without any of the ratio information, colour discrimination would not exist. Am I wrong? Would genuinely love to hear from you on this.
15 comments
[ 2.5 ms ] story [ 45.5 ms ] threadAn alternative title to this article could be "how spaceless numbers achieve defining space"? Like you just have a basis, and this defines space. Mind blown. Really, how do these magnets work?
No. Paint is not light, and has physical properties that constrain what you can do with them. Some colors just don't mix well together, so you'll find recommendation on what exact type of pigment work best for specific mixes.
Perceivable colors are a three-dimensional space, and colors you can make out of three primaries are a three-dimensional space, but they're different shapes. Mixing primary colors is a way of interpolating between them linearly, and the space of perceivable colors doesn't have convenient linear edges.
Here's an example. Your printer has cyan, magenta, and yellow ink (and also black for convenience). So it can print any color, right? It should be able to print bright lime green, something that looks like RGB #00ff00, right?
No, in fact, you can't print that color without "spot color" ink. When you mix cyan and yellow to make green, it will necessarily get darker. If you use enough cyan and yellow to get full-saturation green, it'll be too dark. If you use less ink, it won't be saturated enough.
Lime green is outside of the CMYK gamut. Similarly, the deep cyan you'd get by printing with lots of cyan ink is outside of the RGB gamut.
Where did I try to make this point? My issue is with the quality, not the subject. As pluteoid remarked, this is not a very good article, and fairly long-winded with regard to its content.
There has been many submissions of articles regarding color theory and its relation to psychology, design and physics, many of which are very interesting, and make you learn useful and applicable knowledge. For instance, you may learn how to write an imagemagick script to get the dichromat version of an image.
If each of the values of a three-dimensional coordinate pair encoded frequency, that triplet could carry information about a mixture of up to three colors. This would require sensors that are not "color blind".
I think that's a good summary of the article's point.
I thought printers typically use CMYK (4 inks) or CMYKOG (6 inks).
OP is incorrect; no tristimulus based (RGB or CMY) combination can reproduce the spectral locus without resorting to imaginary colour primaries. Imaginary primaries cannot be reproduced in reality, and as such, “all the colours we see” is “all the colours we may see from the printer.”
If the errors pointed out by above posters are valid, they should really go over to the OP's blog and post them so she can correct the post.
1) The RGB palette used in monitors most certainly does not span the whole gamut of human vision. Not even the wide-gamut monitors used professionally can do that yet.
2) The diagram showing each cone type's peak spectral sensitivity is a bit misleading, too. The M- and L-cones are colored red and green, but if you notice carefully, their peaks lie in the green/yellow area, so characterizing them as red/green is misleading. And "each of our cones seeing a primary colour (blue, green, or red)" is flat-out wrong. Every cone fires in proportion to its overlap. The author clarifies this later, but it's best to tell people the caveats up front. Continually referring to "primary" colors is adding to the confusion here.
3) WhIle they did cite an article pointing out that ratio-computing cells exist in the retina, it does NOT follow that "it is the main source of information that the brain has to help it identify the wavelengths of light". The paper itself only talks about spectral separation in retinal cells pre-brain. I don't know where the author got the idea that ratios are the main source of info sent to the brain. Amplitude information is sent, as well.