Automatic construction of a colormap around the value zero

9
votes

I often want to plot a difference image that ranges from some negative number to some positive one. The issue with that is that matlab's colormaps scale only from the min value to the max value but do not care about some middle value (this case zero). For example, I have produced the peaks image with a single pixel different:

enter image description here

where the original looks like this:

enter image description here

So zero values are blue at the top image and green at the bottom... (using jet colormap as an example)

I'd like to built a colormap that will automatically represent zero values as the same color no matter what will be the min\max values. How can I do that?

3
matlabplotcolor-mapping
I think you can achieve that by scaling everything to a fixed interval prior to plotting, such as [-1,1]. - Marc Claesen
use the colormapeditor: mathworks.de/de/help/matlab/ref/colormapeditor.html there is explained how to focus green at zero - Enno
Robocob specifically asked for an automatic implementation so he wouldn't need to open the colormap editor each time... - bla
you could use caxis to control the color mapping by using a zero-centered range, that way 0 is always in the middle (assuming you are using scaled color mapping as opposed to direct mapping) - Amro

3 Answers

4
votes

you can use this tool from the FEX. It generates blue-white-red colormap, with zero always corresponds to white color. For the case you want to include more than two colors, dig into the code and adapt it to your needs.

There are a few more options in the FEX that do similar things (here's one, and another one)... enter image description here

9
votes

The color mapping can be controlled using the caxis function (which sets the CLim property of the current axes object). The assumption of course is that a scaled color mapping is in use, as opposed to direct color mapping (read the CDataMapping property). By using a zero-centered range as input, you can ensure that zero is always in the middle.

Here is an example:

load penny
P = 4*(P./255) - 1;        %# not centered around zero
imagesc(P), axis image off
colormap(lbmap(64,'BrownBlue')), colorbar

skewed_range

Now we adjust the color mapping to make it symmetric around zero:

c = max(abs([min(P(:)),max(P(:))]));
caxis([-c c])

symmetric_range

Note that I am using the Light Bartlein color scheme which was designed as a divergent palette to highlight differences between two extremes.

If the original range is very skewed, you might want to resample the colormap to give it more values in the sub-range that is over-stretched.

3
votes

I like the other answers, especially the one by @Amro. But it looks so complicated :) And your caxis parameters don't need to depend on your input, you can simply grab them from the axis. All this can be a one-liner:

a = gca; % or whatever you use to access your axis handle
a.CLim = max(abs(a.CLim)) * [-1 1];

That's it. This works in MATLAB R2014b onwards, which has the new figure handle system (see more details here). In earlier versions (and new ones, if you prefer), you would use

a = gca; % or whatever you use to access your axis handle
set(a, 'CLim', max(abs(get(a, 'CLim'))) * [-1 1]);

or, using caxis,

a = gca; % or whatever you use to access your axis handle
caxis(a, max(abs(caxis(a))) * [-1 1]);

Automating this is fairly easy as well:

a = gca; % or whatever you use to access your axis handle
addlistener(a, 'CLimMode', 'PostSet', @(h, e) caxis(e.AffectedObject, max(abs(caxis(e.AffectedObject))) * [-1 1]));

This allows you to do this

for ii = -5 : 5
    imagesc([-1 0 10^ii]);
    pause(0.2);
end

and the middle stripe (representing 0) is always the middle color of the colorbar.