matplotlib image shows in black and white, but I wanted gray

Question

I have a small code sample to plot images in matplotlib, and the image is shown as this :

Notice the image in the black box has black background, while my desired output is this :

My code to plot the image is this :

plt.subplot(111)
plt.imshow(np.abs(img), cmap = 'gray')
plt.title('Level 0'), plt.xticks([]), plt.yticks([])
plt.show()

My understanding is that cmap=grey should display it in grayscale. Below is a snippet of the matrix img being plotted :

[[ 192.77504036 +1.21392817e-11j  151.92357434 +1.21278246e-11j
   140.67585733 +6.71014111e-12j  167.76903747 +2.92050743e-12j
   147.59664180 +2.33718944e-12j   98.27986577 +3.56896094e-12j
    96.16252035 +5.31530804e-12j  112.39194666 +5.86689097e-12j....

What am I missing here ?

Answer 1

The problem seems to be that you have three channels while there should be only one, and that the data should be normalized between [0, 1] . I get a proper looking gray scaled image using this:

import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np

img = mpimg.imread('Lenna.png')
# The formula below can be changed -- the point is that you go from 3 values to 1
imgplot = plt.imshow(np.dot(img[...,:3], [0.33, 0.33, 0.33]), cmap='gray')
plt.show()

This gives me:

Also, a snapshot of the data:

[[ 0.63152942  0.63152942  0.63800002 ...,  0.64705883  0.59658825 0.50341177]
 [ 0.63152942  0.63152942  0.63800002 ...,  0.64705883  0.59658825 0.50341177]
 [ 0.63152942  0.63152942  0.63800002 ...,  0.64705883  0.59658825 0.50341177]
 ...]

Answer 2

For my case, the color (gray) which I wanted is actually "negative" pixels. Subtracting 128 from the image matrix brings the range of pixels from 0-255 to -128 to +127. The negative pixels is displayed in the "gray" color by the package matplotlib.

val = np.subtract(imageMatrix,128)
plt.subplot('111')
plt.imshow(np.abs(val), cmap=plt.get_cmap('gray'),vmin=0,vmax=255)
plt.title('Image'), plt.xticks([]), plt.yticks([])
plt.show()

I will mark my own answer as accepted, as the answer accepted earlier does not talk about treating the pixels on negative scale.