With the macosx backend, images and patches are shown on screen with different color rendering. The Qt backend does work fine. The following script displays an image of a grayscale (np.array) overlaid at the center with the same grayscale generated from matplotlib.patches.

There is a difference of gamma between the image and the patches on what's displayed on screen but if I save this figure, then the resulting file is correct. Checking with some graphic software, the patches are displaying the correct color, not the image.

Snapshot:

Saving the figure from the figure window 'save' button gives this:

Here's the code to test:

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle

color = np.zeros(3)

f, ax = plt.subplots()
im = np.zeros((1000, 1100,3))
for i in range(11):
    im[:,i*100:(i+1)*100,:] = np.tile(np.array([[color]]), (1000, 100, 1))
    ax.add_patch(Rectangle((i*100,300), 100, 400, linewidth=0, facecolor=color))
    color = color + 0.1
plt.imshow(im)
plt.show()