colormap for relative 2d histogram

I have two 2d histograms (in the form of nxm matrices whose elements are populated with 0 or a positive real number). Call them A and B.

I want to overlay A and B to see how they differ, by subtracting so C=A-B, somewhat like in a gene expression map where upregulation is green and downregulation is red. I believe there is a nice colormap available for this already in the File Exchange.

However what I want is somewhat different. If A(25,25)=10.2 and B(25,25)=10.1, so C(25,25)=0.1. Let's also say A(2,1)=0.2 and B(2,1)=0.1, so that C(2,1)=0.1.

I don't want C(25,25) to end up the same color as C(2,1) because I would be losing the information that in both histograms (25,25) is well-populated and (2,1) is not.

One (somewhat still dissatisfactory) solution would be to map the green/red hue of each cell to the sign of the difference A-B so anything positive is green and anything negative is red; and to map the grayscale intensity of the cell to the sum A+B, with white being A+B=0 and black corresponding to some high value.

Does anyone know how to go about doing something like this? It should be simple for someone who knows, but it seems impossible for someone (me) who doesn't! Thanks!

Answer 1

[Walter Roberson]

How about if you plotted the two histograms on the same axis, with different translucent colors? The place where they were the same would be dark because it would have both colors, the one that was greater would extend further up in its own color.

thick = 0.75;  %alpha
bh1 = bar(histogram1);
hold on
bh2 = bar(histogram2);
patches = findobj([bh1, bh2],'-property','AlphaData')
for ph = patches
 set(ph,'AlphaData', thick * get(ph,'AlphaData'));
end

[Rory Staunton]

Thanks Walter that's a good idea, I will try it out!

As of now I was using pcolor. Can you do bar() with a matrix? Which command would be used to actually make the plot?

[Rory Staunton]

(they're 2d histograms) I guess hist3 might work but I have neevr used it...

[Walter Roberson]

There is bar3(). The code I gave should work for bar3() as well.