Hello you all. I am working on a mapping project and need to find the 26 most absolutely differentiable colors - colors that if placed next to one another are the most easily distinguishable.
I tried just dividing the hue slider in Photoshop by 26 and sampling each incrementally but that gave me no where near the flexibility and lightness to darkness contrasts I needed.
Does anyone know of a color format where the whole range of hues and brightness are along the same path so that it is more easily divisible … or perhaps there is a formula somewhere that would allow me to calculate the ‘n’ most differentiable colors? Maybe for the purposes of plotting?
My friend Cindy Brewer used a National Science Foundation grant to research the best color sequences for mapping, and makes the results available as ColorBrewer. But 26 is beyond the scope of choropleth-type mapping.
Tell me more about the mapping project, or seek our advice on the mapping forum CartoTalk. For instance, I can’t imagine needing 26 colors the same intensity. Usually we want some foreground colors and some background colors of subtantially less intensity. Nine dark intense colors, nine grayed colors, and nine pastels, for instance, gets you to 27—but you’d want to use them for very different phenomena.
I guess you want a 3x3x3 slicing of a cubic version of the color solid (like this or these). That should give you 27 different colors/shades: not quite optimal but near enough. I don’t know how to do that in practice though.
I do not know quite what you are trying to achieve here, but FWIW I also think that, however you do this, your map is not going to be at all easily readable by eye. It is often difficult to tell which color from the legend a certain region is meant to be when the number of colors is far lower than 26. It is not that one cannot distinguish two adjacent regions of slightly different color. That is fairly easy. However, if the colors are not all very different, it can be difficult to judge whether two non-adjacent regions are identically colored or just similar.
Actually I really didn’t explain that as well as I could have - what I meant to say was that with the way I was doing it I wasn’t getting the variation in intensity that would help them to be more distinguishable - I know it would be a lot easier to achieve with light and dark variations.
I have done a decent amount of graphing work in my life. I think no matter how you do it - 26 different colors simply isn’t plausible. Sure you can come up with a scheme that will differentiate them as much as possible, but relying on 26 colors to make sense of your data isn’t practical.
If you simply are doing this for display purposes (have no intent of color being assigned to a key) that is one thing, but I think even 12 colors - while doable - is pushing it.
I mean think of resistor color codes - you have 10 there - maybe you can add pink - but what color can you really add without getting confusing? Sure you can break the blues/greys whatever up, but the person has to constantly look to see “is this the dark blue or light blue”? Is this gold or yellow?
For the sake of clarity I am not developing a carto-related map. I am helping a friend ‘map’ unique colors to each letter of the alphabet for a software project.
Do they have to be just color differences. Can you have
red solid
red dots
red horizontal strips
red vertical stripes
red diagonal strips
those five should be easy to distinguish and you could do the same thing for 5 different colors and ad one more.
Or could you do
red circle
red square
red triangle up
red triangle down
red diamond
etc.
Should the viewer just be able to distinguish the colors, or is it also important to be able to see detail between them? That is to say: Black and blue, for instance, are easily discernible, and most folks won’t mistake them for each other. But if you put black text on a blue background (or vice-versa), it’s very difficult to read. So, would they be an acceptable pair for your purposes?
For all practical purposes this would work - any reading of the colors by viewers would be on a white background.
As someone mentioned earlier though, two similar colors while discernible close to one another would work out fine, but if the other color happened to be nowhere around it may be a challenge to tell the difference between light grey and dark grey, for instance.
That’s why I have been trying to hit this with some kind of math in order to shave off as much of that similarity as I can.
This is possibly only a little less naive than the first scheme you had, but what if you do something like this:
Start with the 26 colours round the hue wheel, which are presumably all of similar saturation and brightness. Then
1st one - double saturation
2nd one - keep same
3rd one - halve saturation
4th one - double brightness
5th one - halve brightness
…and so on and so on.
Does that do anything, or is this something you’ve tried already?
That is definitely a viable option. I’ve seen some plotting functions that people have used where you can select ‘n’ colors and the software will generate that many dissimilar colors, but I’ve yet to find a good one and may have to resort to some good old fashioned logical method like this one.
Ignoring color-blind people, an approach would be to start with something like the CIE Lab* or Luv* spaces – where coordinates are based on human perceptual differences, not linearity. Then build a tetrahedron from the 2D gamut; assign colors in staggered fashion from slices of that tetrahedron, e.g. 1-3-5-8-5-3-1. You can Google to find best discussion of CIE Lab*. Here’s a page which may address your question directly.
Since 26 is a large number of distinguishable colors, I’ll guess you won’t even try to support color-blind people – even though 8% of Northern European males suffer from red-green vision deficiency. In an earlier post I described the 7 colors I chose as a compromise for both normal and red-green deficient vision.
The inverse of what you’re looki ng for are the least distinguishable colors, which are contained in MacAdam Ellipses: MacAdam ellipse - Wikipedia
David MacAdam was a professor at the Institute of Optics and a past president of the Optical Society of America who did a huge amount of work on color perception. ( A very modest guy who roamed the Institute halls deressed in a sweater, like an Optical Mister Rogers) His papers are worth reading. Obviously, you want colors whose Macadam Ellipses are far apart if they’re to be easily distinguishable.
Oh… I assumed it was a map project. And got all excited, because I’d just been explaining to my kids all about Topology (and how it addresses ‘How Many Colors Does My Map Need?’).
If your designer friend wants a fun sidenote on colors, why not map a color to its initial letter? So the “Greenish-Blue” T could be Teal or Turquoise.
Let’s see, A is for Amber, B is for Blue… Wikipedia has a list.
I’d start with one of the suggestions above, make some examples with sentences or paragraphs using all the letters (The quick brown fox …), and see which letters need to be tweaked.
One issue I haven’t seen mentioned is that you may need a different set of colors for viewing on screen and for printing. Those will look different, and you could get a set you like, print it, and have it look bad. Even with separate colors for screen and printing, it may vary from monitor to monitor, or from one printer to the next.
