Hacker News new | past | comments | ask | show | jobs | submit login
Impossible color (wikipedia.org)
245 points by sumnulu on July 11, 2018 | hide | past | favorite | 57 comments

I’m surprised the section towards them end about alternating high contrast colors producing impossible colors doesn’t link to “fechner colors” [0]. I am red-green colorblind, but when I was young I would stare at my dark ceiling fan spinning over the white ceiling and see very vivid greens and purples swimming across the blades.

[0]: https://en.m.wikipedia.org/wiki/Fechner_color

Anecdotally, in my first year of university a tutor of mine did a demonstration on Fechner Colour. We walked into the lab and he had a spinning disc on a motor, that was apparently pink in colour. He asked the class (*first year students) what colours we thought each half of the disc were to create the pink shade. Most people guessed pink and white, while a couple of people had correct guesses..

When he stopped the motor, the disc spun down and it was yellow on one side, and blue on the other.

Seems like you're well positioned to add that to the article‽

The interrobang is spreading all over hackernews!

It was the subject of the latest 99 Percent Invisible podcast.

From the people who brought you uptalking? and the Indian Head Wobble, comes another thing which means absolutely nothing.

Just because you don't, like, literally know what it means, doesn't mean it means nothing! In any case, I could care less.

You probably meant "I couldn't care less" - that is, there is nothing I care less about.

My local science centre has had a Benham's top that you spin yourself in one of its permanent exhibits for probably over 15 years now and it's always really cool to see. It fascinates me that you can see yellow, purple, and green on a disk of only black and white.

Wow, I often as a child (still now) would see vivid purple and greens when I closed my eyes or in a dark room. I never thought about it but I wonder what the cause is.

When adjusting paint colors if what you have is too red “or on the red side” you can correct it by adding green. If what you have is on the blue side add yellow then kind of straighten out the undesirable green by adding a smidge of red. This is the sort of thing a skilled shader technician would do to match automotive or sign paint. I did this for a while for PPG

You might find playing with the LAB color scale in photoshop an interesting toy to see why a reddish green color has no sense. by adding red to green you end up on the gray side. basically the colors just neutralize each other.

The thing though (if I got the article right) is that these "reddish green" and "bluish yellow" colours are "virtual". It is almost like they're the result of a glitch in the brain. So I guess standard colour mixing rules don't apply.

Isn’t brown just a reddish green. Or a greenish red.

Yes and no. Brown is just dark orange, which you can sort of get by mixing red and green paints. Paint mixing is weird, though. (I’ve written paint-matching software professionally.)

Sorta, except when dark orange is dark orange and not brown. Your brain will decide between them based on context. The shaded half of a (fruit) orange may well have the same spectrum as the lighted tree trunk it's hanging on, but you will see two different colors.

In substractive color theory, yes, you’ll end up with a brown when mixing red and green in near equal parts.

Keep in mind that the red and green you’ll be mixing is not 100% color pigments, so it cannot merge into a dark grey.

Can you elaborate on why your don't get grey? That was always what I expected and I was surprised that you just get a muddy brown if you mix random paints.

One practical reason you don’t get gray is that pigments have to be real chemicals and so you can’t have a green that perfectly balances a given red’s spectrum. For transparent inks on a white background, if their spectra did match perfectly, you would get a perfect gray. But those perfect dyes don’t exist (or are expensive), hence the K in the CMYK color model.

For opaque paint, it’s weirder: paint scatters and absorbs, so white light goes into the paint layer and starts bouncing around. The resulting reflectance spectrum you see is a function of the ratio of absorption and scattering across the spectrum. Mixing paint mixes the absorption and scattering spectra. One weird result is that adding white can increase a paint’s saturation. (Think of adding apparently-black blue food coloring to white frosting.) Weirder still is you can have two identical-looking paints that, mixed with a third paint make different shades depending on which you use. As a related weird example, yellow plus black can give you a blueish-gray, since the black paint may absorb yellow more than blue (so long as it doesn’t scatter much that’ll still look black) and the yellow will scatter some blue and not absorb all blue (it’s not perfect). So when you mix those you can get something that scatters blue more than it absorbs it and absorbs yellow more than it scatters it. But pick a different black and you’ll get a different result! Color theory is tricky stuff; paint mixing is perhaps the least intuitive part of it.

Thanks. This reminds me of the part of "Surely You're Joking, Mr. Feynman" where a painter says you can mix white and red to get yellow and Feynman says it's impossible and it turns out the painter would use a little bit of yellow to "sharpen it up" [1]. Any way that is possible without yellow paint?

[1] http://blog.everydayscientist.com/wp-content/uploads/feynman...

Interesting. Without very odd paints, I can’t see how you’d get “yellow”, but if your red is desaturated (scattering and not totally absorbing some blue through yellow), and if your white is a bit gray, but your white absorbs and scatters red more than it absorbs and scatters the rest of the spectrum, the mixing path through Lab* color space would (I think) go from red toward light orange, toward warm light gray, finally coming into the light-gray color from the orange direction. I’m a little rusty, but I think that’s right.

Nitpick: what you are calling "saturation" corresponds better to the biconic color space coordinate most people call "chroma". Very dark blues may not be perceptively far from black, but they can have very lopsided spectra and be highly "saturated" in the sense of distance from the gray point in the LUV chromaticity plane.

Basically the counterintuitivity you're citing goes away when you talk about color distinct from luminance, which is how most people here understand it anyway.

The best explanation I've found is in the online book "The Dimensions of Colour" by David Briggs. In particular the section on "Colour Mixing in Paints" (http://www.huevaluechroma.com/061.php). You might have to read some of the preceding sections to get a complete understanding.

The website https://www.handprint.com/ is also very good.

In practice it is possible to mix a grey from coloured paints, but you need disproportionately much blue. Brown is just dark orange, and blue is the complementary colour to orange, so you need to move in the blue direction if you want to get to grey. Also, if you're using red, yellow and blue, you generally need less yellow than red. Try a ratio of 1:2:4 = R:Y:B.

Read the article.

I have always wondered why there is a transition between violet and red, if they are on opposite ends of the frequency spectrum. Despite that, they don't act like two extremes, just two positions on a "wheel". Why is color so wheel-like when the physics is just linear? What are we "seeing" when we see a violet-red? Why do we feel like violet is between blue & red?

There are three kinds of cones in the eye, and each frequency of light is absorbed at a different rate by each of them. So when your eyes see light of a given frequency your brain gets this information as three numbers, which you can visualise as points in 3-dimensional space. One of the directions in this space just represents brightness, so we can throw it away if we're only thinking about colour.

So the different colours are all points in a 2-dimensional space. Each frequency corresponds to a given point in this space. If you plot all the frequencies from the spectrum they form a curve in this space, which is a semicircular shape. Red is at one end of the arc, and violet is at the other.

But the possible colour aren't just the ones on this arc. We can also see colours that correspond to mixtures of different frequencies. These fill in the inside of the semicircle, creating the full space of possible colours, which looks like this (https://pbbhandarkar.files.wordpress.com/2016/10/colorspace....). The spectral colours are on the arc around the outside, and all the colours on the interior can only be made by mixtures. The line joining red and violet is called the "line of purples". It joins violet back up to red to make a circle, but it consists only of mixtures rather than spectral colours. The spectral colours don't go around in a circle, just a semicircle. We need the line of purples to join it back up.

Red receptors are somewhat sensitive in violet part of spectrum too. In standardized response curves it is omitted (I am not sure why), more details: https://midimagic.sgc-hosting.com/huvision.htm

My completely made up intuitive understanding of it is that it's the same as musical pitches; there are many versions of each named pitch e.g A through G that are each an "octave" away from each other. The difference between "Middle C" and the one above it is that the frequency of the higher version is exactly double that of the lower. As a listener the two notes sound extremely similar, the higher one just feels more "energetic" or something. I imagine colors would be the same if our eyes could see more than one "octave" but we are only sensitive to approximately one "octave" of light, ~ 480Thz-750Thz.

Given that, violet appears to be between red and blue in the same way the pitch of G appears to be between the pitches F and A.

Good guess, but completely wrong I'm afraid.

Interesting! Although I don't think this means I'm completely wrong, just that there are "true/full violet" values that we can't see that would make the semi-circle that is visible light into a full circle if we could see a full "octave"; instead we are limited to seeing just violets that are mixes of other colors.

Playing off the "chimerical color" examples - it's possible to approximate the effect of "imaginary colors" (activation of only one cone cell type) by fatiguing the other two and then looking at the target color. e.g. staring for a while at #FF00FF and then switching immediately to a #00FF00 makes the green "pop" in a strange way.

For anyone looking to save skimming through the page:


I get intensely vivid 'stygian blue' hypnagogic hallucinations right before I fall asleep. It's an intensely dark/bright blue. Very hard to describe. It usually manifests as a blob of color that pulses from the outer-edges inwards.

Humorously, Negativland, the amazing decades old plunderphonics group, had a whole prank site[1] and a Over the Edge[2] radio show[3] regarding the 'fourth primary color', Squant. This was also the only color with it's own smell as well. Negativland also had a plugin for web-browser that allowed you to 'see' this magical fourth primary color.




Isn’t it trivial to show impossible colors in VR by texturing an object in different colors like blue and yellow for each eye’s perspective?

Some interesting work around this research paper:

Zhang, Haimo, Xiang Cao, and Shengdong Zhao. "Beyond stereo: an exploration of unconventional binocular presentation for novel visual experience." Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 2012.


Oh wow, these options are really cool. Some of them are hard to see, but even the experience trying is interesting.

My vision seems to oscillate between both eyes in the example picture shown (the yellow-blue one) - in the overlay, the colour changes from yellow to blue and back, with an overlay inbetween. If I concentrate I can see yellow over a blue background consistently though.

Does anyone have the same or a very different experience?

I imagine something similar would happen with your proposal, and possibly make people sick.

(Mixing vision between eyes is really interesting. My eyes are really different (one is good for close-up, one better for distance) so I have some experience with this.)

For me it is blurred yellow-blue striped tape that goes slowly from left to right (idk why, but since you mentioned, my right eye is a little ‘for close-up’). Sometimes one eye takes over and I see just one of the colors.

Traditional displays (such as what would be used in vr goggles) do not produce yellow light. You only get reddish-green with rgb colors

Technically you could make such an extended gamut OLED screen, 4-stimulus. Probably RBGY variant. Some such screens are on the market already but they do not really produce superior results. Some phone screens use RBGW pattern too.

Not really, if you look at the CIE chromaticity diagram, you’ll find that the experience of spectral yellows are very closely approximated by an RGB monitor. It’s spectral cyan that is really out of gamut for a standard display. Take a photo of a green traffic light (which is a little cyan so colorblind people can see it’s not red). I’m pretty sure that green-light color is out of gamut for most displays. https://en.wikipedia.org/wiki/SRGB

Your comment inspired me to look at my phone under a microscope (I opened a paint program's color picker).

As with many things I have examined under a microscope, I didn't actually learn anything new, but it was still interesting to see for myself.

Crap, guess we’re stuck with reddish green :(

Artist Sterling Crispin has a VR painting program [1] that explores this allowing you to paint different colors to each eye simultaneously.

[1] http://www.cyberpaintapp.com/

Reminds me of The Color Out of Space.


Reminds me of Pratchett's Octarine. And seems reasonable to assume that impossible colours are only visible to wizards and cats.

Or "ulfire" and "jale" from "A Voyage to Arcturus" by David Lindsay.

What happens when someone with tetrachromacy looks at these same test colour tests?

See: https://en.wikipedia.org/wiki/Tetrachromacy

Still looking for neon brown.

You joke, but that’s a great example: brown is really dark orange; neon is fluorescent, and so brighter than your eye thinks should be possible for reflection, so sadly, no neon poo emoji.

> Impossible colors or forbidden colors are supposed colors that cannot be perceived in normal seeing of light that is a combination of various intensities of the various frequencies of visible light, but are reported to be seen in special circumstances.

This article could use some serious copy editing.

So do it then.

It would probably be better if somebody who understands the article does it.

Now I want to visit Manchester to see if Reddish Green is also impossible

That explains the invisible pink unicorn and colorless green dreams sleeping furiously. Flying spaghetti monster is still a mystery.

Is this something to do the THAT DRESS?

No, it's a different bug in the human optical system.

Impossible colors are derived from the frequency-response curve of the three standard pigments in the retina. (There are people with 2 pigments and thus have a reduced color space, and people with 4 pigments who have an increased color space.) Since the pigments are not evenly distributed, the brain synthesizes some colors from incomplete information.

The dress problem stems from luminance-color correction in the brain itself. The eye has quite an amazing dynamic contrast ratio, but a much narrower static contrast ratio. If you have picked up on clues that the photograph was over-exposed in one way, your brain interprets that as a signal that the colors are washed out. If you have picked up on clues that the photograph is undersaturated, your brain says that the colors should be brighter than the pixels are. The photograph in question gave ambiguous clues.


Nah. Not Han Solo coat on Hoth, either.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact