Matplotlib.Colors.LogNorm problems with zeros in 2D-plots when saving to eps

Question

In my python program I plot some 2D graphs with imshow:

img=imshow(data,
           interpolation='none',
           vmin=0.0001,
           vmax=0.01,
           cmap=custom_map,
           norm=LogNorm()
           )

When I save the plot as *.eps or *.pdf I get some strange lines in the pictures, where the values of data are very close to vmin.

When I save the plot as *.png or *.svg or *.jpg this error doesn't occur.

How can I avoid that python draws this strange lines?

enter image description here

EDIT: Here is a minimal code which should reproduce the problem. At least I get the error all the times...

from matplotlib.colors import LogNorm
import numpy as np
import matplotlib.pyplot as pl

A=np.array([[0.000000,0.000426],[0.000000,0.000524]], dtype=np.float)
pl.imshow(A, interpolation='none', norm=LogNorm())

pl.savefig('test.eps')
pl.savefig('test.png')

Even if you do not use LogNorm() but use np.log(A) the problem occurs in the output.

Edit 2 With

A=np.array([[np.log(0),np.log(0),np.log(0.1),np.log(0.000426)],
        [np.log(0),np.log(0),np.log(0),np.log(0.000524)]], dtype=np.float)

the problem occurs only at the borders from np.log(0) to some real values as you can see in the next picture

enter image description here

does your custom map have alpha in it? See https://github.com/matplotlib/matplotlib/issues/2473. From the code you gave here there is no way to reproduce your problem (and hence sort out what your problem is). Please post a _minimal_ amount of code which will reproduce the problem. _Most_ matplotlib problems can be reproduced with < 15 LoC. — tacaswell, May 27 '14 at 14:03
To be honest, I can't reproduce the problem with any analyticval function which should fit to the shown pattern.... I tried several things: No `vmin` and `vmax`, no `LogNorm`, no `cmap`... the problem occurs only when saving to .eps and .pdf. When the picture is opened in `spyder` after the `pl.show()`-line, there are no black lines. So the error occurs when saving the file!? EDIT: my custom map has no alpha in it. — FrankTheTank, May 27 '14 at 14:50
The thing you see on the screen is probably rendered through Agg, the ps and pdf paths use two entirely different (vector based) code paths (backends). This is why getting a simple reproducible example is important. Also, what version of mpl are you using? — tacaswell, May 27 '14 at 15:07
Thank you for your answers. I added a piece of code with which my problem is reproducable... I use version 1.3.1 of mpl and PyCharm as IDE. — FrankTheTank, May 28 '14 at 07:52

score 0 · Answer 1 · edited May 23 '17 at 12:20

Seems like this is a problem with LogNorm() and the zeros in the array... So the bad pixels should have a special colorvalue which can be manually set. For the people who are looking for answers: here is a description: How can I plot NaN values as a special color with imshow in matplotlib?

Now change the code in the opening post:

cmap = matplotlib.cm.jet
cmap.set_bad('w',1.)
pl.imshow(A, interpolation='none', cmap=cmap, norm=LogNorm())

With this the eps output is correct - or better: as wanted.

If there are some other ideas, please add them to the comments or answers.

enter image description here

Matplotlib.Colors.LogNorm problems with zeros in 2D-plots when saving to eps

1 Answers1