Question

I’m trying to create a 2d histogram that overlays three separate datasets. My idea is to color the datasets red, green, and blue, so that the density of the red data in a bin corresponds to the red value of that bin’s color, and likewise for blue and green.

There are examples of this, but so far as I can tell there’s no implementation of polychromatic plotting in matplotlib.

Getting to a grid of the form

[[ (r,g,b) , (r,g,b) ... (r,g,b) ] ,
 [ (r,g,b) , (r,g,b) ...         ] , 
   .......                   
 [ (r,g,b) , (r,g,b) ... (r,g,b) ]]

is no problem at all. The problem is that all the plotting functions I’ve found want to map single bin values onto a color scale, they don’t allow me to set the full color value of the bins.

Is there some primitive I should be looking for? Is there already a histogram to do this?

Answer 1

If your data is already in an RGB format (weighted between 0 and 1), then imshow will interpret a (M,N,3)-shaped array as an RGB array. Use interpolation='nearest' to pixelise the output:

import numpy as np
import matplotlib.pyplot as plt

# Example pixel array
pixels = np.random.rand(100,100,3)
reds = pixels[:,:,1:] = 0.          # Remove B and G

plt.imshow(reds, interpolation='nearest')

If it is stored as a list of tuples, then a np.array(...) will do the conversion.