6

I'm trying to render an image at its true dimensions (not scaled or stretched) and the easiest way to do this with matplotlib seems to be figimage.

However, when I try to use it in a Jupyter notebook, the figure doesn't show. Other plots show fine, this only seems to affect figimage:

figimage not showing

As you can see, this first plot shows fine, but the second one does not. What am I doing wrong?

When I run the following code in an IPython shell , the figure shows up as expected, so maybe it's a problem with my Jupyter setup?

import matplotlib
from matplotlib import pyplot as plt
import numpy as np

x = np.linspace(0, 2*np.pi, 500)
plt.plot(x, np.sin(x))
plt.show()

data = np.random.random((500,500))
plt.figimage(data)
plt.show()
nfelger
  • 823
  • 9
  • 21

1 Answers1

6

figimage only adds a background to the current figure. If you don't have an already existing figure, the command wont render anything. The following snippet will work both inside and outside IPython Notebook:

%matplotlib inline
import matplotlib
import matplotlib.pyplot as plt
import numpy as np

plt.figure()

x = np.linspace(0, 2 * np.pi, 500)
plt.plot(x, np.sin(x))

data = np.random.randn(500, 500)
plt.figimage(data)

plt.show()

However, it doesn't do what you want/expect. In order to render an image in its true dimensions you would have to play with figsize and dpi, as others have attempted previously [1] [2] [3] [4]:

data = np.random.randn(500, 500)
dpi = 80
shape = data.shape

fig, ax = plt.subplots(figsize=(shape[1]/float(dpi), shape[0]/float(dpi)), dpi=dpi, frameon=False)
ax.imshow(data, extent=(0,1,1,0))
ax.set_xticks([])  # remove xticks
ax.set_yticks([])  # remove yticks
ax.axis('off')     # hide axis
fig.subplots_adjust(bottom=0, top=1, left=0, right=1, wspace=0, hspace=0)  # streches the image and removes margins
fig.savefig('/tmp/random.png', dpi=dpi, pad_inches=0, transparent=True) # Optional: save figure
fig.show()
Imanol Luengo
  • 15,366
  • 2
  • 49
  • 67
  • Oh, I see. `figimage` is primarily for rendering backgrounds then? Your suggestion works a charm (although I don't fully understand the need for the subplots), thanks! – nfelger Sep 18 '17 at 15:38
  • 1
    @nfelger Exactly, the short answer is that `figimage` is just for backgrounds. As for the subplots, there are not completely necessary, you could replace all those commands with `plt.xticks([])` (and similar for other commands), but I like coding with `fig` and `axes` as I think it makes the code more clear, particularly when you have multiple subplots. If you replace the first command with `plt.figure(.......)` it does exactly the same. – Imanol Luengo Sep 18 '17 at 16:43