使用 PIL/Pillow 使用调色板识别图像的颜色值

Question

I'm trying to identify the colors of the used color palette of an image with PIL/pillow. I've tried the following:

• image[x,y] : this will only give me the index number of the corresponding pixel (ie 1 )
• image.getpixel((x,y)) : again, this will only give me the index number of the corresponding pixel (ie 1 )
• image.getcolors() : This will give me the number of pixels and their corresponding index number (ie [(2, 1), (2, 0)] )
• image.palette : Returns a "PIL.ImagePalette.ImagePalette object"
• image.getpalette() : Returns a large array of (to me seemingly) unrelated integers (ie [0, 0, 255, 255, 0, 0, 2, 2, 2, 3, 3 ,3 ...] )

As an absolute fallback, I could convert the image mode and then get the color values, but I'd rather not if possible.

With this example image (2x2 pixel image, indexed mode with 2 colors created with GIMP, the top two pixels are red (255,0,0) the bottom two are blue (0,0,255)), I was expecting something like:

image.getpalette()
1: (255,0,0)
0: (0,0,255)

---

Edit: The closest I have is:

image.palette.getdata() : This gives me ('RGB;L', b'\\x00\\x00\\xff\\xff\\x00\\x00') . Is there any way to have this mapped to the index number. Here each three bytes would map to one index number, I'd reckon.

Answer 1

You can get and arrange the palette like this:

import numpy as np
from PIL import Image

# Open image
im = Image.open('a.png')

# Get palette and reshape into 3 columns
palette = np.array(im.getpalette(),dtype=np.uint8).reshape((256,3))

Then just print palette .

[[  0   0 255]      <--- first entry is blue
 [255   0   0]      <--- second is red
 [  2   2   2]      <--- grey padding to end
 [  3   3   3]
 [  4   4   4]
 [  5   5   5]
 [  6   6   6]
 ...
 ...
 [253 253 253]
 [254 254 254]
 [255 255 255]]

If you want to count the colours and how many of each there are, do this:

# Convert Image to RGB and make into Numpy array
na = np.array(im.convert('RGB')) 

# Get used colours and counts of each
colours, counts = np.unique(na.reshape(-1,3), axis=0, return_counts=1)    

That gives colours as:

array([[  0,   0, 255],
       [255,   0,   0]], dtype=uint8)

and counts as:

array([2, 2])