Python (or anything else vaguely sensible) - image conversion to monochrome.

[pythonmegapixel]

So here's the deal.

 

I have a folder full of loads of PNG graphics. Does anyone know of a Python library that would let me copy these files, but with every pixel that is not transparent set to black, and every pixel that is transparent remaining transparent?

 

I'm fairly familiar with python, and I'm also happy to use another language if needed, but if you want to link me to a library I'd prefer some pointers on what within the library I should be using (what functions/classes?) rather than just a link and nothing else... I'm not the most familiar with this kind of task.

[AbydosOne]

If you don't mind Java, I could whip something up over lunch (like in a couple hours).

 

The pseudo-code gist:

• Load in image
• Iterate through each pixel, testing pixel 'A' channel value (how do you want to handle partial transparency?)
  • if 'A' is transparent, skip; else, if 'A' is not transparent, write black to pixel
• Write out image

[pythonmegapixel]

9 minutes ago, AbydosOne said:

If you don't mind Java, I could whip something up over lunch (like in a couple hours).

 

The pseudo-code gist:

• Load in image
• Iterate through each pixel, testing pixel 'A' channel value (how do you want to handle partial transparency?)
  • if 'A' is transparent, skip; else, if 'A' is not transparent, write black to pixel
• Write out image

What you've described is indeed exactly what I need.

 

As for the handling of partial transparency... anything other than entirely transparent needs to be black. So essentially if the alpha-channel is anything less than 255, the pixel needs to be black.

 

I have no objection in principle to Java, but I don't have an environment for writing Java set up at the moment, so I would have to do that.

[elpiop]

33 minutes ago, pythonmegapixel said:

What you've described is indeed exactly what I need.

 

As for the handling of partial transparency... anything other than entirely transparent needs to be black. So essentially if the alpha-channel is anything less than 255, the pixel needs to be black.

 

I have no objection in principle to Java, but I don't have an environment for writing Java set up at the moment, so I would have to do that.

You can use the pillow library to do this. 

 

from PIL import Image

im = Image.open('test.png')

for x in range(im.width):
    for y in range(im.height):
        r, g, b, transparency = im.getpixel((x,y))
        if transparency == 255:
            im.putpixel((x,y), (0, 0, 0, 255))

im = im.save('out.png')                        

 

[pythonmegapixel]

Using various responses from this thread and elsewhere, I've managed to cobble together a solution in Python using Pillow. I thought I'd just put my code here incase it helps anyone else:

from PIL import Image
import glob

for f in glob.glob("./png/*.png"):
	img = Image.open(f)
	pixels = img.load()
	size = img.size
	for w in range(size[0]):
		for h in range(size[1]):
			if pixels[w,h][3] > 0: # If the alpha channel is more than zero (i.e. is partially or fully transparent)
				pixels [w,h] = (0, 0, 0, 255) # Set to black, fully opaque
			else:
				pixels [w,h] = (0,0,0,0) # Set to fully transparent
	
	img.save(f + "_BLACK.png")