Python: How to save EXACT numpy array data to image using matplotlib.image.imsave()

Question

I observed some strange behavior when saving and loading data using matplotlibs imsave() and imread() functions. The RGB values I save are different to the ones I get when loading the picture back in again.

import numpy as np
from matplotlib import image

s_pic = np.zeros((1, 1, 3))

s_pic[0,0,0] = 0.123
s_pic[0,0,1] = 0.456
s_pic[0,0,2] = 0.789

image.imsave('pic.png', s_pic)

l_pic = image.imread('pic.png')

print(l_pic[0,0,0])
print(l_pic[0,0,1])
print(l_pic[0,0,2])

The output I get is:

0.12156863
0.45490196
0.7882353

Can somebody explain why the RGB values change in this process? I have checked the matplotlib documentation but couldn't find an answer to this question.

wwii · Accepted Answer · 2021-03-22 16:51:02Z

1

Can somebody explain why the RGB values change in this process?

RGB values are integers in the range 0-255. Your float is interpreted as:

>>> .123 * 255
31.365
>>> int(.123 * 255)
31

Thirty-one is being written to that pixel. Then the reverse..

>>>
>>> 31 / 255
0.12156862745098039
>>>

Delving into the source for imsave() the array passed to imsave() is converted to RGBA values using matplotlib.cm.ScalarMappable().to_rgba(bytes=True)

>>> from matplotlib import cm
>>> sm = cm.ScalarMappable()
>>> rgba = sm.to_rgba(s_pic, bytes=True)
>>> rgba
array([[[ 31, 116, 201, 255]]], dtype=uint8)
>>>

edited Mar 22, 2021 at 16:51

answered Mar 21, 2021 at 21:39

wwii

23.9k7 gold badges42 silver badges81 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Jack the Scripter Over a year ago

Thank you very much! I will use PIL to save and load the image, since it seems that matplotlib doesn't allow me to do what I want in this case.

Collectives™ on Stack Overflow

Python: How to save EXACT numpy array data to image using matplotlib.image.imsave()

1 Answer 1

1 Comment

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Related