Question

I have a 2D array that I’m plotting with imshow and I would like to have costums colors depending on the value of each pixel of my array. I’ll explain it with an example.

from pylab import *
from numpy import *

img = ones((5,5))
img[1][1] = 2

imshow(img,interpolation='nearest');colorbar()

If you ran this code you would see a red square in a blue background. The red square corresponds to the pixel [1][1] in img, while the other pixel are colored blue because they have a value of 1. What if I want the red square to be colored with a custom color?
Or more generally, if I have a 2D array like img in the example, how can I color pixel with the same value with a color I can choose.

I have found this page that explains how to generate a custom colorbar but that’s not useful: http://www.scipy.org/Cookbook/Matplotlib/Show_colormaps

Answer 1

That link you sent has the following:

But, what if I think those colormaps are ugly? Well, just make your
own using matplotlib.colors.LinearSegmentedColormap. First, create a
script that will map the range (0,1) to values in the RGB spectrum. In
this dictionary, you will have a series of tuples for each color
‘red’, ‘green’, and ‘blue’. The first elements in each of these color
series needs to be ordered from 0 to 1, with arbitrary spacing
inbetween. Now, consider (0.5, 1.0, 0.7) in the ‘red’ series below.
This tuple says that at 0.5 in the range from (0,1) , interpolate from
below to 1.0, and above from 0.7. Often, the second two values in each
tuple will be the same, but using diferent values is helpful for
putting breaks in your colormap. This is easier understand than might
sound, as demonstrated by this simple script:

   1 from pylab import *
   2 cdict = {'red': ((0.0, 0.0, 0.0),
   3                  (0.5, 1.0, 0.7),
   4                  (1.0, 1.0, 1.0)),
   5          'green': ((0.0, 0.0, 0.0),
   6                    (0.5, 1.0, 0.0),
   7                    (1.0, 1.0, 1.0)),
   8          'blue': ((0.0, 0.0, 0.0),
   9                   (0.5, 1.0, 0.0),
  10                   (1.0, 0.5, 1.0))}
  11 my_cmap = matplotlib.colors.LinearSegmentedColormap('my_colormap',cdict,256)
  12 pcolor(rand(10,10),cmap=my_cmap)
  13 colorbar()

Isn’t this exactly what you want?

Here’s an example of how to do it with the image you provided:

import matplotlib
from matplotlib import pyplot as plt
from pylab import *

img = ones((5,5))
img[1][1] = 2

cdict = {'red': ((0.0, 0.0, 0.0),
                (0.5, 1.0, 0.7),
                     (1.0, 1.0, 1.0)),
             'green': ((0.0, 0.0, 0.0),
                       (0.5, 1.0, 0.0),
                       (1.0, 1.0, 1.0)),
             'blue': ((0.0, 0.0, 0.0),
                      (0.5, 1.0, 0.0),
                     (1.0, 0.5, 1.0))}

my_cmap = matplotlib.colors.LinearSegmentedColormap('my_colormap',cdict,256)
plt.pcolor(img,cmap=my_cmap)
plt.colorbar()
plt.show()

Also, if you really want to map a number to a colour you can use discrete_cmap as specified in that example you linked to, here’s the example method the scipy documentation provides:

def discrete_cmap(N=8):
    """create a colormap with N (N<15) discrete colors and register it"""
    # define individual colors as hex values
    cpool = [ '#bd2309', '#bbb12d', '#1480fa', '#14fa2f', '#000000',
              '#faf214', '#2edfea', '#ea2ec4', '#ea2e40', '#cdcdcd',
              '#577a4d', '#2e46c0', '#f59422', '#219774', '#8086d9' ]
    cmap3 = col.ListedColormap(cpool[0:N], 'indexed')
    cm.register_cmap(cmap=cmap3)