How to create an image from pixel color values (R, G, B)

Question

I am new to python (also to stackoverflow) and I need help with combining pixel color values into an image.

After searching through the web and playing around with the codes mentioned, I eventually used the following code to get the color values from a given image:

    from PIL import Image  

    filename = "INSERT_PATH_OF_IMAGE"  
    image = Image.open(filename)  

    t = list(image.getdata())  
    r = list(image.getdata(0))  
    g = list(image.getdata(1))  
    b = list(image.getdata(2))  

    print(r)  
    print(g)  
    print(b)  
    print(image)  

This outputs the numerical values of the given pixels of the given image like so:

    [r_value1, r_value2, ..]
    [g_value1, g_value2, ..]
    [b_ , b_..]

How from here would the pixel-coordinates of the colors be identified and how could they be put together into an image?

How can the print-command be structured such that

    [(r_value1, g_value1, b_value1), 
    (r_value2, g_value2, ..)]

would be the output shown?

Because (depending on image size) the amount of values shown can exceed the output terminal record (Using Pycharm), how could the printed result instead be saved in a text file? (Like: Editor on Windows OS)

The eventual goal is to create an algorithm to use with opencv (probably) with which to process images, I thought I should start with understanding how to process just standard images from pixeldata first and progress from there.

Thanks for reading!

P.S. I welcome any and every advice on what sources I should read up on, so feel free to share. =)

Answer 1

As mentioned in the comments, try to avoid Python lists of image pixels - they are slow and inefficient. Prefer instead Numpy arrays and vectorised processing by Numpy or OpenCV.

from PIL import Image
import Numpy as np

# Make a linear gradient and resize to 8x8 so we can see it
gradient = Image.linear_gradient('L').resize((8,8))

# Rotate to make interesting
gradient90 = gradient.rotate(90)

# Make a solid black image same size
black = Image.new('L', (8,8))

# Merge the 3 channels into RGB
rgb = Image.merge('RGB',(gradient, gradient90, black))
rgb.show()

# Make into Numpy array
na = np.array(rgb)

# Reshape as 8x8x3 RGB array into 64x3 triplets
triplets = na.reshape((-1,3))
print(triplets)

[[ 16  16   0]
 [ 16  47   0]
 [ 16  80   0]
 [ 16 112   0]
 [ 16 144   0]
 [ 16 176   0]
 [ 16 208   0]
 [ 16 239   0]
 [ 47  16   0]
 [ 47  47   0]
 [ 47  80   0]
 [ 47 112   0]
 [ 47 144   0]
 [ 47 176   0]
 [ 47 208   0]
 [ 47 239   0]
 [ 80  16   0]
 [ 80  47   0]
 [ 80  80   0]
 [ 80 112   0]
 [ 80 144   0]
 [ 80 176   0]
 [ 80 208   0]
 [ 80 239   0]
 [112  16   0]
 [112  47   0]
 [112  80   0]
 [112 112   0]
 [112 144   0]
 [112 176   0]
 ...
 ...
 [239 208   0]
 [239 239   0]]

---

# Save as CSV
np.savetxt('image.csv', triplets, fmt='%i', delimiter=',')

16,16,0
16,47,0
16,80,0
16,112,0
16,144,0
16,176,0
16,208,0
16,239,0
47,16,0
47,47,0
47,80,0
...
...

---

If you want to output the CSV to your PyCharm terminal, i.e. stdout, you can use:

import sys

np.savetxt(sys.stdout, triplets, fmt='%i', delimiter=',')

---

If you want to load the image back from the CSV, you will need to know its original shape because that is not stored in the CSV. Note that it is generally a poor idea to store an image in a CSV because it is neither efficient, nor compressible, nor viewable.

It would go like this:

# Load pixels from CSV into Numpy array and reshape to original shape
pixels = np.loadtxt('image.csv', delimiter=',', dtype=np.uint8).reshape((8,8,3))

# Convert Numpy array into PIL Image
im = Image.fromarray(pixels)

# Display on screen
im.show()

# Save to disk
im.save('result.png')