在 geoTIFF 中获取所有独特颜色时，如何知道 PIL 指的是什么颜色？

Question

I have a bunch of geoTIFF images I am trying to analyze. For example, here is one below:

I would like to obtain a count of how many pixels in this image are blue, and how many are white. I am using the following code to obtain this information:

from PIL import Image
from collections import defaultdict

# open the image (fp is the filepath of the image)
im = Image.open(fp)
# the image is in 'P' mode

# convert to RGB?            
# im.convert('RGB')
            
by_color = defaultdict(int)
for pixel in im.getdata():
    by_color[pixel] += 1

im_col = Image.Image.getcolors(im)
print(im_col)
print(by_color)

These images are in 'P', or palettized mode. Whether or not I convert these images to 'RGB' mode by uncommenting that line of code above, I get the following output in this format (this specific output is for the example image above):

[(777, 0), (196, 1), (378, 2), (1149, 3)]
defaultdict(<class 'int'>, {2: 378, 3: 1149, 1: 196, 0: 777})

So, from my understanding, the 0, 1, 2, and 3 represent the unique pixel colors, and the respective 777, 196, 378, and 1149 represent the amount of times these pixels are in the image (and if I add up 777 + 196 + 378 + 1149 = 2500, this makes sense, since this is a 50 pixel x 50 pixel image).

My two main concerns are:

1. Why are there four values? According to the image, there should only be three values: black background, blue ocean, and white ice - is there a hidden color somewhere?
2. How can I know what colors the 0, 1, 2, and 3 represent in palletized format? I looked for some color code on some website, and I found that 0 meant 2, but the 1 meant green, which I know isn't right, so I'm definitely misunderstanding something here. Is there a way to get this 'color code' from the image using PIL?

Thanks in advance.

Answer 1

Your image is a little bit weird, it has two pairs of identical palette entries:

• index 1 == index 3, and
• index 2 == index 4

---

You can see that with ImageMagick if you run this command in Terminal:

magick identify -verbose 1QgAe.png

Abbreviated output

...
...
  Histogram:
    12431: (0,0,0) #000000 black
    6205: (0,0,254) #0000FE srgb(0,0,254)
    21364: (254,254,223) #FEFEDF srgb(254,254,223)
  Colormap entries: 256
  Colormap:
    0: (0,0,0,1) #000000FF black
    1: (254,254,223,1) #FEFEDFFF srgba(254,254,223,1)
    2: (0,0,254,1) #0000FEFF srgba(0,0,254,1)
    3: (254,254,223,1) #FEFEDFFF srgba(254,254,223,1)
    4: (0,0,254,1) #0000FEFF srgba(0,0,254,1)
...
...

The 21,364 pixels that are srgb(254,254,223) are just the combined totals of 3,126 and 18,238.

---

You can equally see it with pngcheck like this:

pngcheck -p 1QgAe.png 
File: 1QgAe.png (6135 bytes)
PLTE chunk: 256 palette entries
  0:  (  0,  0,  0) = (0x00,0x00,0x00)
  1:  (254,254,223) = (0xfe,0xfe,0xdf)
  2:  (  0,  0,254) = (0x00,0x00,0xfe)
  3:  (254,254,223) = (0xfe,0xfe,0xdf)
  4:  (  0,  0,254) = (0x00,0x00,0xfe)
  5:  (  0,  0,  0) = (0x00,0x00,0x00)
  ...
  ...

---

If you want to see that in Python with PIL, you can use this code and you will see it is the same:

#!/usr/bin/env python3

from PIL import Image

# Load image
im = Image.open('1QgAe.png')

if im.mode=='P':
    colours = im.getcolors()
    print(f'Colours: {colours}')

    palette = im.getpalette()
    for i in range(256):
        r = palette[3*i]
        g = palette[3*i+1]
        b = palette[3*i+2]
        print(f'palette[{i}] = rgb({r},{g},{b})')

Sample output

Colours: [(12431, 0), (3126, 1), (6205, 2), (18238, 3)]

palette[0] = rgb(0,0,0)
palette[1] = rgb(254,254,223)
palette[2] = rgb(0,0,254)
palette[3] = rgb(254,254,223)
palette[4] = rgb(0,0,254)
palette[5] = rgb(0,0,0)
palette[6] = rgb(0,0,0)

If you are comfortable with Numpy, you can use:

from PIL import Image
import numpy as np

# Load image
im = Image.open('1QgAe.png')
palette = np.array(im.getpalette()).reshape(256,3)
print(palette)