SimScale CAE Forum

Rainbow versus 'Cool to Warm' color mappings


#1

It has been a while since I read the latest ParaView tutorial by Kenneth Moreland…

I now see that since version 5.2, there has been a rather strong recommendation that rainbow color mappings never be used in visualizations

I for, one have just gotten used to the drawbacks of the rainbow color mapping and have never really thought much about it. I did not even know there were drawbacks.

It turns out there is large number of studies which confirm that data is obfuscated when using the rainbow mapping.

You will find a summary of a few of the issues in Chapter 2 starting at pg 48…

I was a true believer by the time I got to the bottom image you see here:

And then I tried changing a few of my latest mappings to the ‘Cool to Warm’ mapping. I must say, that I see much more detail than with Rainbow …

Give the section a read…

NOTE: This is not about a promotion to use ParaView (although I think you should put ParaView in your data analysis arsenal), it is relevant for all data presentations…


#2

Have heard the same from other people as well that the rainbow filter should not be used, thanks for pointing this out dale!

Also feel free to give a source to the picture you posted, thanks!

Jousef


#3

I recently published an article where the reviewer recommand the use of a different colormap (not the rainbow colormap) saying that " Figures should be reformatted to still give the detailed information when it print in black and white format ".

So, I used a new reformatted color map that appears to be perceptually-uniform color map and the results appear more meaningful.

NOTE : I used MATLAB for the postprocessing.


#4

Well after trying to use ParaView’s ‘Cool to Warm’ (blue to dark brown) color scheme for the last couple weeks, I now see that I will have room for both approaches in my presentations.

I am convinced that virtually all of the drawbacks of Rainbow schemes can be alleviated for my type of color distributions (where there are whole ‘contourlike’ distributions of colors in my images) with the use of a discretized Rainbow color map (which I have been doing for as long as I can remember in my SimScale career ).

By that I mean, the Rainbow scheme drawbacks mainly appear when a continuous color map is used.

If you use a Rainbow color map that is composed of a limited number of colors so that contour lines ‘appear’ in the image, there should be no obfuscation of data. In fact, in some of my images, I can use 20 or more colors and still see these contour lines well enough to take any area on the image of a certain color and determine its value range on the color scale.

I have not been able to discretize the ParaView cool to warm scheme to more than a few contour lines before it becomes impossible to determine any areas’ value range using discrete color ranges.

I would venture to say the a DISCRETIZED Rainbow color scheme is MUCH better at data presentation than ‘Cool to Warm’ color scheme in some images.

Here is a simple (very few color contour image, I will post one with more discrete colors soon):

I just came across this image where I used 18 colors (over and under colors included). The ParaView LIC visualization contours do obfuscate some of the color contours somewhat, but even still, by counting level changes from an obvious color range , you can still isolate the value range of any of the 18 colors in the scale: