Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
Math

A New Study Overturns 100-Year-Old Understanding of Color Perception (phys.org) 67

Posted by BeauHD from the unseen-errors dept.
An anonymous reader quotes a report from Phys.Org: A new study corrects an important error in the 3D mathematical space developed by the Nobel Prize-winning physicist Erwin Schrodinger and others, and used by scientists and industry for more than 100 years to describe how your eye distinguishes one color from another. The research has the potential to boost scientific data visualizations, improve TVs and recalibrate the textile and paint industries. [...] "Our original idea was to develop algorithms to automatically improve color maps for data visualization, to make them easier to understand and interpret," [said Roxana Bujack, a computer scientist with a background in mathematics who creates scientific visualizations at Los Alamos National Laboratory and lead author of the paper]. So the team was surprised when they discovered they were the first to determine that the longstanding application of Riemannian geometry, which allows generalizing straight lines to curved surfaces, didn't work.

To create industry standards, a precise mathematical model of perceived color space is needed. First attempts used Euclidean spaces -- the familiar geometry taught in many high schools; more advanced models used Riemannian geometry. The models plot red, green and blue in the 3D space. Those are the colors registered most strongly by light-detecting cones on our retinas, and -- not surprisingly -- the colors that blend to create all the images on your RGB computer screen. In the study, which blends psychology, biology and mathematics, Bujack and her colleagues discovered that using Riemannian geometry overestimates the perception of large color differences. That's because people perceive a big difference in color to be less than the sum you would get if you added up small differences in color that lie between two widely separated shades. Riemannian geometry cannot account for this effect. "We didn't expect this, and we don't know the exact geometry of this new color space yet," Bujack said. "We might be able to think of it normally but with an added dampening or weighing function that pulls long distances in, making them shorter. But we can't prove it yet."

The findings appear in the journal Proceedings of the National Academy of Science.
This discussion has been archived. No new comments can be posted.

A New Study Overturns 100-Year-Old Understanding of Color Perception More

A New Study Overturns 100-Year-Old Understanding of Color Perception

Comments Filter:

  • I thought perceptual image compression codecs already took into account that human eyes are pretty much crap for distinguishing subtle differences in color? It seems like they've just discovered the same effect, just under a different area of research.

    • Re: (Score:2)

      by Luckyo ( 1726890 )

      We also knew this from evolutionary biology side for a long time, as color perception evolved to primarily enable fruit eaters to differentiate optimally ripe fruit from one that is suboptimal for consumption.

      There's very little need for excellent perception of minute differences in this process, so we don't have that ability as it wasn't selected for. Just like many carnivores don't have the ability to see color as we do at all, because they don't need it as they do not consume fruit.

      • I had seen some science program putting forth the idea that perception of color was also culturally influenced.

        You already have some cultures that don't distinguish between two colors that are at least linguistically distinct here, but they could differentiate between very subtle differences in the shade of blue (for instance).

        Never heard anything more about it, so take it with a grain of salt.

        • Re: (Score:3, Interesting)

          by DarnOregonian ( 10088276 )

          You already have some cultures that don't distinguish between two colors that are at least linguistically distinct here, but they could differentiate between very subtle differences in the shade of blue (for instance).

          That's before you consider tetrachromacy [sciencealert.com]. Some women are born with four types of cones instead of three and therefore can perceive a lot more colors.

          • Is that the kind of receptors that allow some women to claim that "Ballet" is an actual color that is required for bridesmaid dresses?

            I still insist that "peach" and "lime" are fruit and not a colors.

            • Is that the kind of receptors that allow some women to claim that "Ballet" is an actual color that is required for bridesmaid dresses?

              No but it does explain why they have to have so many vials of red nail polish.

              I still insist that "peach" and "lime" are fruit and not a colors.

              "The color "orange" is named for the fruit, [etymonline.com] not the other way around. So, calling a shade of green "lime" or a shade of pink "peach" is not unprecedented.

          • We're not sure if they perceive more colors, but they would perceive them differently. Note that our three types of cones aren't spread out evenly across the spectrum (otherwise we'd have Red, Green, Violet). People that have this extra type of cone haven't reported that there is this amazing color that no one ever talks about and that there's no word for. Cones don't see in binary, so there's a lot of overlap even with the three cones, having a fourth cone that's not in the infrared or ultraviolet range

          • Actually, it's even MORE subtle than that.

            Technically, everyone has one more pseudo-color receptor than normally gets counted: rods.

            Rods have a peak sensitivity that approximately corresponds to "blue-green". Under EXACTLY the right conditions, you can even perceive it... but it's so subtle, unless you actually KNOW what you're seeing, you probably won't even RECOGNIZE it as the presence of a fourth primary color cue. You're most likely to perceive it in a situation where your dark-adapted, but suddenly fin

        • Not just cultural. Gender [thedoghousediaries.com] also matters.

        • Re: (Score:2)

          by znrt ( 2424692 )

          it's because "cultures" and "languages" all impose slightly different instances of categorical thinking. here's an excellent lecture about exactly this:

          https://www.youtube.com/watch?... [youtube.com]

      • Re: (Score:2)

        by znrt ( 2424692 )

        We also knew this from evolutionary biology side for a long time

        no we didn't? high precision is probably an edge case but color perception isn't only an advantage for fruit eaters, many predators can locate suitable prey or can avoid toxic prey because of color and maybe even more importantly color is decisive for mate selection in many species (you know, that bizarre thingy where fit genes get actually passed on, which is how evolution works).

        • Re: (Score:2)

          by Luckyo ( 1726890 )

          Color based selection is limited to what species can otherwise see for self evident reason of "if you can't see it, it's pointless".

          That said, there are some cases where specific coloration can function in multiple parts of the spectrum which is a side effect rather than actual sought effect. Humans skin color comes to mind, as our UV reflectivity adaptation also grants skin the absorbing effect at higher wavelengths. People who have the skin that looks whiter in UV spectrum look darker in visible spectrum.

    • human eyes are pretty much crap for distinguishing subtle differences in color?

      It really depends on context. If you have ever tried to make a subtle gradient, our eyes are annoyingly good at seeing the subtle seems in the gradient.

      • If you want to confirm that, I found a quite relaxing game:

        http://i-love-hue.com/ [i-love-hue.com]

        It is about sorting colored tiles into a smooth (and surprisingly satisfying) color gradient. It's interesting to see that you absolutely can't compare similar colors in a slightly different context, but in a gradient can instinctively see where the smoothness is disturbed.

        (Disclaimer: Not in any way related to game, author or publisher)

      • A similar problem is the fact that there are actually TWO major variants of "green" and "red" cones commonly found in men with "normal" color perception whose relative sensitivity peaks are a few nm apart. Let's call them G1 and G2, and R1 and R2.

        Men with R1G1, or R2G2, tend to make aesthetic color choices that are mutually acceptable to both each other and to most women, because hue-perception is RELATIVE, and R1G1 and R2G2 have approximately the same distance between G and R.

        The problems occur with R1G2 a

    • The opposite: This study concludes that if you ask people how different two colors are, the perceived difference is smaller than the sum of the subtle differences between colors that are in-between. We seem to perceptually "exaggerate" subtle differences.

    • Re:While TFS is a bit above my pay grade (Score:5, Informative)

      by thegarbz ( 1787294 ) on Friday August 12, 2022 @03:39AM (#62782756)

      They didn't "discover" this effect, they identified the existing method of quantifying it is incorrect. The potential implication of this is that it may one day lead to even better quality image compression.

      Though in reality perceptual based image compression isn't actually anywhere near as finely tuned as any colour theory, so it's unlikely to make any meaningful difference in this regard.

      A more likely outcome is that it may give us new tools for measuring the quality of display rather than the standard DeltaE measurement of perceptual difference we currently use.

      • Compression standards like JPEG and MPEG already implement a crude but effective type of "perceptual image compression" by converting the image to YUV color space, and halving the resolution of the U,V color components since we're far less sensitive to color than brightness (Y component).

        I guess in theory you could more heavily compress specific parts of the color spectrum that we're less sensitive to, but I rather doubt it would provide enough benefit to convince appliance makers to support a new standard.

        • Yeah that's what I meant with "isn't anywhere near as finely tuned as colour theory". You could dramatically increase complexity and reduce efficiency chasing only slight improvements over what we already have.

          but I rather doubt it would provide enough benefit to convince appliance makers to support a new standard.

          Remember when JPEG2000 was going to revolutionise the PC? Oh man ... I just realised I'm getting old.

    • I read it that they didn't discover that effect, but found a mathematical representation that behaves similar so it reduces all the "fudging" that is needed to take those effects into account.

      This is a reminder that sensory input is never linear, and (as in audio) not even strictly logarithmic, so there still is potential for models that represent that "beautifully"

      Which of course my end up in "simple" functions, but in dimensional spaces that are based in really really fancy math...

    • This seems to be more about human inability to judge relative distances between colors.

  • New color space? (Score:4, Funny)

    by CaptQuark ( 2706165 ) on Friday August 12, 2022 @12:42AM (#62782560)

    Oh, great. Another new color space for Photoshop to map items to. Photoshop color spaces [photoshopessentials.com]

    • "CIE 1931 oughtta be enough for anybody!"

    • The Photoshop color spaces don't even matter. As long as colors on a monitor are created by a combination of RGB pixels, those will be the entirety of your color space.

      If we could somehow make monitors out of oil paints, that would give us a different color space to work with. But we can't. We're stuck with the serious limitations of RGB.

      • Re:New color space? (Score:5, Informative)

        by thegarbz ( 1787294 ) on Friday August 12, 2022 @03:34AM (#62782748)

        The Photoshop color spaces don't even matter. As long as colors on a monitor are created by a combination of RGB pixels, those will be the entirety of your color space.

        Absolutely false. The monitor determines the maximum limits of the monitor's ability to produce something. A colour space is created from that and used to translate colour information for accurate display on the screen.

        If your monitor's colour space is larger than sRGB (which many on the market are), then working in sRGB in Photoshop becomes a limiting factor. The colour space used in Photoshop (or working space used in any application) absolutely does matter, in fact the *smallest* colour space in any part of the storage > editing > display chain always matters.

        RGB doesn't actually have very serious limitations. Colour spaces like REC-2020 cover nearly every colour we can see, and more critically for recording purposes most of nature and what we see actually falls within what we have in sRGB. These days the only thing colours we really can't display properly in an RGB format is violet and some very specific shades of ultra saturated greens. The rest is already so close that the limitations of RGB are basically imperceptible.

        • Re: (Score:2)

          by Shinobi ( 19308 )

          Bright/vivid orange shades are also difficult in sRGB. Muted orange can be ok, which is why photos of oranges often have lighting that sorts of mutes it, but bright/vivid oranges lose a lot of nuance, though it used to be even worse back in analogue TV/video days.

          • There's a lot of limitations with sRGB, but I think what phantomfive was saying was that through the use of Red Green and Blue is a significant limitation. Only on a scientific level though. Have a look at oranges on an OLED panel with a REC-2020 colour space and you'll see we can get amazingly bright and vivid colours using three primaries.

            Actually orange shades aren't that much of a challenge in sRGB either, the blue/green transition is. A nice deep cyan and green is the most immediately obvious differenc

      • You're missing the point by reducing the Photoshops(*) to the digital world of monitors. But images created digitally are eventually ending up in the physical world. Usually printing. Then you already are outside RGB and even outside additive color mixing. You even have to transform from a 3D into 4D (CMYK) space.

        Add Pantone colors or special effects outside the regular color range (metallic effects, night glow...)

        It is because that monitors are limited to RGB these additional color spaces matter.

        And there

    • Not yet. This is only telling us the tools we used to develop colour spaces may not account accurately for our perception of colour. It'll be along time before this has any impact on the tools we use.

  • Because these mathematicians ignored Lettvin (Score:5, Informative)

    by Anonymous Coward on Friday August 12, 2022 @01:21AM (#62782602)

    Jerry Lettvin published the neurological basis of color vision in roughly 1977, based on his earlier work applying electrodes to single neurons.

            http://jerome.lettvin.com/jero... [lettvin.com]

    Color vision is not based on direct color intensity: it's based on edge detection, because light entensity and light source matter so much, but to identify objects as being the same object as seen previously it's far more effective, and easier to evolve, to compare and contrast the object with its surroundings. The phenomenon is fundamental to many, many optical illusions.

    This can be overwhelmed by other issues, such as only the more red sensitive "rods" of the retina being activated in low light levels, or people with only two rather than three dyes in their eye cells being insensitive to red-green differences. But if the vision modelers didn't pay attention to Lettvin's work, well, their foolishness has led them astray for decades.

    • Re: (Score:2)

      by MehNo ( 9614184 )
      I am having some edge detection issues while reading that poorly scanned paper.

    • Re:Because these mathematicians ignored Lettvin (Score:4, Interesting)

      by hnwombat ( 172691 ) on Friday August 12, 2022 @07:29AM (#62783044)

      Not only that, but we don't directly perceive red, green, and blue. What we actually perceive DIRECTLY is a dark indigo ("blue", or short-wave photopigment), green ("green", or medium-wave photpigment), and yellow-green ("red", or long-wave photopigment). These correspond to about 420nM, 534nM, and 564nM respectively, with peak overall response at about 555nM. All the rest of the gamut are tricks your brain figures out. Which is why things like language, training, etc., etc., can affect our perception of "color".

      So you don't see red ever, or very much blue. You construct what you think those are from indigo, green, and yellow-green.

      And we're not even going to go into mutants with more or fewer photopigments. Or mantis shrimp, which have somewhere around 12-16 different photopigments. Proof that they are aliens.

      • Not only that, but we don't directly perceive red, green, and blue. What we actually perceive DIRECTLY is a dark indigo ("blue", or short-wave photopigment), green ("green", or medium-wave photpigment), and yellow-green ("red", or long-wave photopigment).

        Our cones responses overlap, so we generally see a mix no matter what the frequency of a single visible EM wave. I would guess that our actual defined primary colors are historical/environmental and could be slightly shifted without much difference.

        T

    • Proof that color is not solely based on intensity can be seen on Page 4 [slideplayer.com] as demonstrated by that picture of a burger.

      How can we see multiple colors where there are only two colors [slideplayer.com]? The rest of the presentation [slideplayer.com] goes into detail on how Edwin H. Land (of Polaroid fame) discovered that the difference in frequency is what mattered and called it Retinex Theory. [wikipedia.org] It is an incomplete theory in that it doesn't explain optical illusions [wikipedia.org] such as that checkerboard pattern where both the light and dark squares are the s

  • In other words ... (Score:4, Interesting)

    by fahrbot-bot ( 874524 ) on Friday August 12, 2022 @02:26AM (#62782654)

    The scientists thought The Dress [wikipedia.org] was white and gold too.

  • This is well-known way before this paper as far back as last decade (or before). See the work in BT.2100 / BT.2124 per Dolby in ICtCp. The trick is to find the geometry and transformation (manageable) into computational models. Actually, the deltaE measure is good enough for SDR.

  • So this is a negative result paper, without any attempt of proposing something better. Weak.

    • And? Identifying something is being applied incorrectly is weak because you don't have all the answers? What of your post? It provided nothing at all of value to anyone anywhere. Are you racing the people who conducted the study to the bottom? Is your goal to prove you are the weakest?

      • Apparently you cannot differentiate the expected standards for a Slashdot comment and those for a scientific article. In serious scientific journals, where a majority of submitted articles are rejected, such articles are rejected on the ground of "insignificant results" not advancing sufficiently the field.

  • You insensitive clod

    • Re: (Score:3)

      by godrik ( 1287354 )

      I'm color blind as well. And I am excited about any new understanding of color perception.

      I started to wear enchroma glasses last week. And they are really nice. If we can understand better how we see colors, then maybe we will get better color correction glasses in the future!

      So thanks to these scientists for their work!

      • I had to look up enchroma glasses, and I'm glad I did, I learnt something. Thank you. I'm not colour-blind, and I know those glasses only work for some types of colour-blindness, but when it does, the results are awesome. See for instance this video [youtube.com].
  • > That's because people perceive a big difference in color to be less than the sum you would get if you added up small differences in color that lie between two widely separated shades.

    Related to Weber-Fechner?

    https://en.wikipedia.org/wiki/Weber%E2%80%93Fechner_law

  • There is no "error in Riemann geometry." (Score:4, Interesting)

    by caveat ( 26803 ) on Friday August 12, 2022 @09:21AM (#62783304)

    When I read the summary, and the article, I was wondering how showing an error in Riemannian geometry would only be a "oh, our model of color perception is wrong" rather than a massively fundamental change in physics, given that Riemann's techniques gave Einstein the tools to whip up General Relativity - an error in those rules would be infinitely more significant than mere color spaces.

    No no, once you get to the actual paper, it merely turns out that color perception can't be adequately described by a Riemann manifold.- "We show that a Riemannian metric overestimates the perception of large color differences because large color differences are perceived as less than the sum of small differences. This effect, called diminishing returns, cannot exist in a Riemannian geometry."

      Interesting, and definitely useful, but there's no flaw in the underlying mathematical framework. The usual sensationalist bullshit.

    • I'm not sure how you got that understanding from the summary, but the summary makes no claim about an error in Riemann geometry. It makes a claim about "corrects an important error in the 3D mathematical space" and that error was the Riemann geometry was applied in a situation where it shouldn't have been.

      In fact let's review some key phrases in the summary:
      determine that the longstanding application of Riemannian geometry
      using Riemannian geometry overestimates the perception of large color differences
      Ri

  • ... that straight guys will have to learn the names of twelve different shades of blue?

  • So, they determined that an old hack was not as good as some had thought. So what?

  • The study found that a widely used distance metric in a perceptual color space model is inaccurate, especially over long distances. Calling it a "math error" as phys.org does, is clickbaity. "All models are wrong, but some are useful." The current models have brought us quite far, but they could certainly be improved.

Slashdot Top Deals

A morsel of genuine history is a thing so rare as to be always valuable. -- Thomas Jefferson

Close