È meglio usare numpy
oltre a PIL per eseguire la matematica delle singole bande di un'immagine.
Come un esempio inventato che è non destinati a guardare bene in qualsiasi modo:
import Image
import numpy as np
im = Image.open('snapshot.jpg')
# In this case, it's a 3-band (red, green, blue) image
# so we'll unpack the bands into 3 separate 2D arrays.
r, g, b = np.array(im).T
# Let's make an alpha (transparency) band based on where blue is < 100
a = np.zeros_like(b)
a[b < 100] = 255
# Random math... This isn't meant to look good...
# Keep in mind that these are unsigned 8-bit integers, and will overflow.
# You may want to convert to floats for some calculations.
r = (b + g) * 5
# Put things back together and save the result...
im = Image.fromarray(np.dstack([item.T for item in (r,g,b,a)]))
im.save('output.png')
ingresso
uscita
fonte
2012-05-31 01:34:16
bene grande grazie per l'esempio – clifgray