1

I'm looking for a color space that gives me a color wheel where two pairs of hues each with the same number of degrees apart gives the same perceived distance of hue.

I found this for the CIELAB color space:

enter image description here

This is much better than what I get with the simple HSL/HSV spaces, but I still think it's a bit off: I can hardly tell 28 and 27 apart, but on the opposite side, 100 and 90, is very different.

My use case isn't compression but to chose colors for categories in a user interface. Their numbers can vary and I need good control over their perceived similarity.

  • That is not "color space". That is a color wheel. – Rafael Oct 1 '18 at 8:09
  • Hence the title of this question. If you've got a wheel that's not derived from a color space, I'd take that too. – John Oct 1 '18 at 8:10
  • 1
    Maybe LCH, maybe some muinsell space? But realistically speaking this is one fo the real problems of colorwheel like thinking is that once you do make it uniform distance you end up making certain colors less important thna others, whch in turn will make your complementary color selection patterns invalid. – joojaa Oct 1 '18 at 13:17
  • Check-out the work by Faber Birren who based all of his colour research on the human perception of colour rather than their technological relationships numerically. All his experiments were made with actual colour specimens and human visualization. – Stan Oct 1 '18 at 22:41
2

Your image isn't CIELab color wheel. Chroma values (=sqrt(a^2+b^2)) and the lightnesses are random, they seem like they are selected for good looks. Chroma and lightness should be the same in all samples AND also possible to be shown right in sRGB screens (see NOTE1) Otherwise the adjacent samples have random perceptual distances.

Conclusion: Do not use this image as an evidence which proves that CIELab is not peceptually uniform.

NOTE1: There are much colors in CIELab, which are impossible for normal computer screens (=out of the sRGB gamut)

I suggest you read the following: http://scanline.ca/hue/cielab.html as a first aid.

CIELab gives to you good tools for making such circle because it is designed to be perceptually uniform. Only have a stripe of color samples with uniform hue spacing, equal chroma and equal lightness. You must use chroma and lightness values which fit in all hues to sRGB range. At least Chroma=40% of max and Lightness=70% of max should be ok.

In Photoshop's Lab color system you cannot input Hue nor Chroma values directly. GIMP has HCL which is a polar coordinate version of HCL.

I prepared the following stripe of samples in Photoshop's Lab color system. The hues are from 0 to 360 degrees, step is 15 degrees. Lightness is 70% of max and chroma is 40% of 127. The first and the last sample are the same a=51, b=0

enter image description here

The values are calculated in the following table:

enter image description here

The greyed columns are only because sin and cos need radians and I wanted to see the real chroma when a and b were rounded to integers.

The formulas:

a = 0,01 * chroma% * 127 * cos (Hue)

b = 0,01 * chroma% * 127 * sin (Hue)

Chroma values of the samples can be increased without stepping out of the sRGB range, if the lightness are reduced. Here's another stripe with higher chroma:

enter image description here

  • Ok, but my goal isn't to get CIELAB. My goal is to have a better, more subjectively uniform color wheel. – John Oct 1 '18 at 12:07
  • @John I added the material to one constant hue step color circle. All samples have the same chromaticity and the same lightness. I used CIELab (=Lab in Photoshop). – user287001 Oct 1 '18 at 19:18
  • Ok, I will play around with Gimp later, thanks for all the input. But just to check with a different pair of eyes: The 2-3 middle blues in your series still look closer together than any other neighbours, right? – John Oct 2 '18 at 8:09
  • @John those stripes have gone through quite a bunch of non-color managed format conversions since they were created as PSD in my computer. The colors can be on your screen be toally different. All blues seem different to me, but the longer I watch then, the more they start to look equal. Greens and reds have the same effect. Our color perception isn't reliable, everything depends on what is around. – user287001 Oct 3 '18 at 8:14
1

1. You can not...

Perception is a physiological thing. You would need to change how human vision works. We simply have more sensitivity to shifts on Green tones, then on Red ones and we have little sensitivity to blue shifts. See it by yourself.

A color wheel is a methodological approach to predict numerically some values.

2. Ok, yes you can use some other "color wheels"

The approach that aims to order more perceptually based the colors is the Munsell Color System. It is not a wheel, it is a solid.

Or make your own

Using a vector based program where you can control radial gradients.

1

A Rational Colour Circle

There are tweaks of the 12 hue colour wheel originally envisioned by Michel Chevreul called "rational" colour circles that compensate for the failings of digital displays of pigments.

Birren explains his colour-circle in accordance with the practicalities of art and artists. He differentiates between «warm» and «cold» colours, and recommends a «Rational Color Circle» which groups 13 colours around a grey which is displaced from the centre. He thus presents a system which assigns more space to the «warm» colours between red and yellow than the «cold» colours occurring between green and violet. As a result, Birren takes account of the eye’s ability to distinguish more warm colours, which are therefore granted greater importance in art.

Colour order systems in art and science

  • Human perception does not have a linear response. It is not subject to the neat and orderly mathematical relationships used to try to diagram it. It is not symmetrical. It is as disorderly as our behaviour is. You must find or devise a compensation factor to accommodate its irregularities for your need. – Stan Oct 1 '18 at 23:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.