python - White lines in matplotlib's pcolor

23
votes

In some pdf viewers such as Preview on OSX, plots made with matplotlib's pcolor have white lines (see image below). How can I get rid of them?

The source code is very simple (choose any data for x,y,z):

import matplotlib
matplotlib.use("pdf")
import matplotlib.pyplot as pl
pl.figure()
pl.pcolormesh(x,y,z,cmap="Blues",linewidth=0) 
pl.savefig("heatmap.pdf")

enter image description here

4
pythonmatplotlib
This is because pcolormesh actually draws individual rectangles, and some pdf viewers render them differently than others. Both as a workaround and because it's more efficient, it's better to use imshow in this case. For example, in your case: plt.imshow(z, interpolation='nearest', cmap='Blues', extent=[x.min(), x.max(), y.min(), y.max()]). - Joe Kington
Thank you! That was exactly what I was locking for. - hanno
You can work around that by passing in rasterized=True to colorbar. - Joe Kington
I'm getting an "unexpected keyword argument 'rasterized'" error when calling pl.colorbar(rasterized=True). With MPL version 1.3.1 and 1.4.x. The top answer on this question works for me. - farenorth
Try cb.solids.set_rasterized(True) - hanno

4 Answers

27
votes

The comments have a good solution that utilizes imshow. When imshow is not appropriate for the input data (e.g. it is not uniformly spaced) this generally solves this problem,

pcol = pl.pcolormesh(x,y,z,cmap="Blues",linewidth=0,)
pcol.set_edgecolor('face')

If that approach does not reduce the lines sufficiently, you can also try this:

pl.pcolormesh(x,y,z,cmap="Blues",linewidth=0,rasterized=True)

In addition to reducing the lines between squares this approach also tends to reduce file size a bit, which is sometimes useful. In this case you may want to tune the dpi setting when saving (e.g. pl.savefig("heatmap.pdf", dpi=300)) until you get something that is satisfactory.

6
votes

The accepted answer didn't work very well for me. I seemed to have gotten closer by using antialiased=True, in addition to linewidth=0. This was with matplotlib version 3.0.2. Note the middle plot corresponds to the best version.

fig, axes = plt.subplots(1,3, figsize=(15,5))
axes[0].pcolormesh(XX, YY, ZZ_r, zorder=-1, norm=norm, cmap='magma', alpha=0.5, antialiased=True)
axes[1].pcolormesh(XX, YY, ZZ_r, zorder=-1, norm=norm, cmap='magma', alpha=0.5, antialiased=True, linewidth=0.0)
axes[2].pcolormesh(XX, YY, ZZ_r, zorder=-1, norm=norm, cmap='magma', alpha=0.5, antialiased=False, linewidth=0.0)

enter image description here

5
votes

I had this problem, using rasterized=True solved it (on matplotlib version 3.1.0).

1
votes

Building on the answer by @eqzx: If you set alpha=1, you will get better results as far as grid lines are concerned. However, your application/use case may differ.

import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0, 10, 100)
y = np.linspace(0, 10, 100)
xx, yy = np.meshgrid(x, y)

zz = xx**2 + yy*2
fig, axes = plt.subplots(2,3, figsize=(15,5))
axes[0,0].pcolormesh(XX, YY, ZZ_r, zorder=-1, cmap='magma', alpha=0.5, antialiased=True)
axes[0,1].pcolormesh(XX, YY, ZZ_r, zorder=-1, cmap='magma', alpha=0.5, antialiased=True, linewidth=0.0)
axes[0,2].pcolormesh(XX, YY, ZZ_r, zorder=-1, cmap='magma', alpha=0.5, antialiased=False, linewidth=0.0)
axes[1,0].pcolormesh(XX, YY, ZZ_r, zorder=-1, cmap='magma', alpha=1, antialiased=True)
axes[1,1].pcolormesh(XX, YY, ZZ_r, zorder=-1, cmap='magma', alpha=1, antialiased=True, linewidth=0.0)
axes[1,2].pcolormesh(XX, YY, ZZ_r, zorder=-1, cmap='magma', alpha=1, antialiased=False, linewidth=0.0)

See plot here