CODEWITHSUNDEEP
pythonnumpypython-imaging-library
Z. Uryum
Z. Uryum

Reputation: 55

Python PIL/numpy conversion

I'm having problems converting between python PIL images and numpy arrays. I already checked existing Stackoverflow posts on this, but it didn't solve the problem:

import matplotlib.pyplot as plt
import numpy as np
import PIL

rgb_img = plt.imread('some-image.png')
PIL_rgb_img = PIL.Image.fromarray(np.uint8(rgb_img))

plt.imshow(PIL_rgb_img)

I get a black screen. I tried with and without converting to uint8, and I also tried only keeping the RGB channels out of the entire RGBA data. Nothing worked.

Upvotes: 2

Views: 2551

Answers (2)

Yasin Yousif
Yasin Yousif

Reputation: 967

I may not give you a full explanation , (for that, you may read matplotlib's functions docs) but clearly with some tests the following is happening:

when you call:

rgb_img = plt.imread('img.png')

it gives a numpy float array, which will read colors between [0 - 1] as white and black ( and for RGB also )

when you call: 

PIL_rgb_img = PIL.Image.fromarray(np.uint8(rgb_img))

which convert it to uint8 values, it just take what supposed to be 255 and make it 1 which is completely wrong,

you know in uint8 the values should be between [0 - 255]

and when you put :

plt.imshow(PIL_rgb_img)

it just show a 255 times 'faded' image , which is very close to black..

P. S.:

That's only happen with '.png' files, something with plt.imread ..

to solve just put :

img = 'some_img.png'
rgb_img = plt.imread(img)
if img.split('.')[-1]=='png':
    PIL_rgb_img = PIL.Image.fromarray(np.uint8(rgb_img*255))
else:
    PIL_rgb_img = PIL.Image.fromarray(np.uint8(rgb_img))

plt.imshow(PIL_rgb_img)

That's should fix it.

Upvotes: 2

Abhishek Mahato
Abhishek Mahato

Reputation: 49

from PIL import Image
import numpy as np

#PIL to Numpy
pil_img = Image.open('some-image.png')
numpy_img = np.asarray(pil_img)

#Numpy to PIL
resultImage = Image.fromarray(numpy_img)

Upvotes: 1

Related Questions