Matplotlib is what you want.
The special sauce is a rendering technology known as agg which i believe stands for anti-grain geometry. The results is that this library produces publication-quality graphics across a staggering range of plot families.
My plot isn't all that much like the one in linked to in your Question, but that's only because the colors are different.
That sort of plot is trivial in Matplotlib. Here's the code i used to create it.
import numpy as NP
from matplotlib import pyplot as MPL
import matplotlib.cm as CM
M = NP.random.randint(0, 25, 100**2).reshape(100, 100)
M = NP.where(M < 20, 0, M)
fig = MPL.figure(frameon='False')
ax1 = fig.add_subplot(111)
ax1.imshow(M, cmap=CM.Accent, interpolation='nearest')