Author Topic: No more rainbows!  (Read 1896 times)

Anette von der Handt

  • Global Moderator
  • Professor
  • *****
  • Posts: 231
    • UMN Probelab
No more rainbows!
« on: December 21, 2017, 03:29:09 pm »
Hi,

I stumbled upon this passionate discussion of the misuse and use of the rainbow colormap that we (or at least I) love so dearly.

https://agilescientific.com/blog/2017/12/14/no-more-rainbows

A view that is shared by many others (apparently there are even anti-rainbow-colormap-activists) and led to a change in the default colormap for Matlab users for example (Parula, which is intellectual property, unfortunately) in 2014.

To sum up some of the critiques:
  • No natural, instinctive ordering of the colors (what comes first, green or yellow?)
  • It can create phantom features: perceived sharp transitions in data that are just sharp transitions between hues
  • The green and cyan sections of the rainbow color map are perceptually indistinct, so it is essentially non-linear
  • Loss of critical information about high and low data values when printing in grey scale
  • Difficult to read for color-blind people

I really like the rainbow colors but the more I learn about data visualization the more I understand its disadvantages but in general. I think its advantage at this point is just our familiarity with this "standard" color transition and that we intuitively understand maps using it. But are we missing the train?

Ben Buse posted earlier in this forum a color pallet for surfer based on Viridis, a colormap that works well for color-blind people.
http://probesoftware.com/smf/index.php?topic=958.0

So, dear probers, should we adopt a new and better "standard" colormap for our element maps? To an extent we are in a good position to make a change happen as we generate the data for these element maps. Which one should it be?


More on the rainbow color map:

More on Parula and Viridis color maps
Against the dark, a tall white fountain played.

Probeman

  • Emeritus
  • *****
  • Posts: 1787
  • Never sleeps...
    • John Donovan
Re: No more rainbows!
« Reply #1 on: December 21, 2017, 04:32:50 pm »
I concur.

This is why I've always preferred the "thermal" color scale.  It's much more intuitive in that black is minimum and white is maximum with brown, red, orange, yellow in between.

In fact I tried hard with the rainbow color scale in PFE and CalcImage, so we don't have black at both sides of the rainbow color scale as seen here:



This is a reasonable compromise I think, though I still prefer the thermal scale as seen here:



Though some people are fans of this color scale used by JEOL:

The only stupid question is the one not asked!

glennpoirier

  • Professor
  • ****
  • Posts: 32
Re: No more rainbows!
« Reply #2 on: December 21, 2017, 04:48:23 pm »
As a token colorblind person I heartily concur with abandoning the rainbow colour map.  It's never made any sense to my eyes. I used to like a pure thermal colour map but honestly I prefer the default thermal map used in PFE, my eyes can pick up very small brightness differences with this LUT. The Viridis LUT is nicer than the rainbow but I'm not very green sensitive.
My two cents
Cheers
Glenn

Anette von der Handt

  • Global Moderator
  • Professor
  • *****
  • Posts: 231
    • UMN Probelab
Re: No more rainbows!
« Reply #3 on: December 21, 2017, 05:00:37 pm »
Interesting, the JEOL one is very similar to Viridis.

Viridis


And here is Parula for comparison
Parula


There are some more contenders:
Magma


Inferno


Plasma



All found here:
http://medvis.org/2016/02/23/better-than-the-rainbow-the-matplotlib-alternative-colormaps/

Of course "Magma" would appeal to the geologists...
Against the dark, a tall white fountain played.

Karsten Goemann

  • Global Moderator
  • Professor
  • *****
  • Posts: 173
Re: No more rainbows!
« Reply #4 on: December 21, 2017, 05:17:22 pm »
My 2 cents regarding thermal would be that it is basically monochrome, so what's the benefit compared to just greyscale?


Anette von der Handt

  • Global Moderator
  • Professor
  • *****
  • Posts: 231
    • UMN Probelab
Re: No more rainbows!
« Reply #5 on: December 21, 2017, 05:25:04 pm »
My 2 cents regarding thermal would be that it is basically monochrome, so what's the benefit compared to just greyscale?



You make the poor, struggling publishers happy by having to pay their fee for having color figures in the paper ;)
Against the dark, a tall white fountain played.

Anette von der Handt

  • Global Moderator
  • Professor
  • *****
  • Posts: 231
    • UMN Probelab
Re: No more rainbows!
« Reply #6 on: December 21, 2017, 05:33:33 pm »
Here is a screenshot from a Youtube video that compares the different colormaps


Left: Jet and Parula; Center: Magma and Inferno; Right: Plasma and Viridis.

Video:
https://www.youtube.com/watch?time_continue=9&v=u9a4NO3iGgA

Based on this, I like Inferno the best.
Against the dark, a tall white fountain played.

Probeman

  • Emeritus
  • *****
  • Posts: 1787
  • Never sleeps...
    • John Donovan
Re: No more rainbows!
« Reply #7 on: December 21, 2017, 05:49:50 pm »
My 2 cents regarding thermal would be that it is basically monochrome, so what's the benefit compared to just greyscale?

My understanding of the benefit of thermal compared to grayscale is that the human eye can only discern some 60 gray levels, while with the thermal scale the adding of red, orange and yellow (to black and white), greatly improves visual sensitivity (millions of colors).

Here is an interesting link discussing these issues.

https://betterfigures.org/2015/06/23/picking-a-colour-scale-for-scientific-graphics/

They make one claim that I find quite strange: "In general, people interpret darker colours as representing “more”."  That's certainly not the way I see it, but then I'm pretty strange...   more photons is lighter, right?   ;D
john
« Last Edit: December 21, 2017, 07:49:22 pm by Probeman »
The only stupid question is the one not asked!

John Donovan

  • Administrator
  • Emeritus
  • *****
  • Posts: 2355
  • Other duties as assigned...
    • Probe Software
Re: No more rainbows!
« Reply #8 on: December 21, 2017, 05:55:04 pm »
Most of the pixels of the strawberries in this image are gray:



https://www.digitalrev.com/article/your-brain-will-colour-correct-this-greyscale-image-of-strawberries

Your brain *wants* to see color!
John J. Donovan, Pres. 
(541) 343-3400

"Not Absolutely Certain, Yet Reliable"