An Image-Pixelizer in Python (with ImageMagick)
A few days ago a friend of mine challenged me with the need for an image pixelizer: given an image, the program has to divide it in nine squares, blurring all but the center one.Image Magick to the rescue!
This has been my very first thought about it. So, I quickly wrote a raw Python script that launched in turn two Image Magick applications:
convert and montage. The former was used to produce nine images with the parts of the original image, the latter to reassemble them after having blurred every one except one.
Before continuing, having a source image as the following one
Pure Python Implementation
This is how I implemented in pure Python (the code is available on my GitHub repository):#!python
import sys
import itertools
import glob
import shlex
import subprocess
from wand.image import Image
# Example of invocation: img_pixelier.py src9.png
if __name__ == '__main__':
src_file = sys.argv[ 1 ]
src_image = Image( filename = src_file )
src_width = src_image.width
src_height = src_image.height
pieces = 3
crop_width = int( src_width / pieces )
crop_height = int( src_height / pieces )
dst_image = Image( width = src_width, height = src_height )
for r in range(0, pieces):
for c in range(0, pieces):
start_at_x = c * crop_width
start_at_y = r * crop_height
segment = Image( filename = src_file )
segment.crop( start_at_x, start_at_y, width = crop_width, height = crop_height )
if c != r or c != int( pieces / 2 ) or r != int( pieces / 2 ):
segment.blur( sigma = 20 )
dst_image.composite( image = segment, left = start_at_x, top = start_at_y )
dst_image.save( filename = 'blurred.png' )
In order to work, there is the need for the
wand package, that is the binding to Image Magick.
The first step is to create an image from the given filename, and to compute the original dimensions of the image.
Then I assume to split the image into pieces blocks per row and colum, so that pieces x pieces parts will be produced, and hence I compute how large every single sub-image is going to be large.
Then I create a dst_image new image of the size of the original image, and loop over rows and columns performing a crop of the original image with adjusted offset depending on the piece I want to crop.
If the current sub-image is not the center one, i.e., r and c are different or not the half of their values, the sub image is also blurred.
At the end of the loop, I save the image on disk so that the final result is produced.
Conclusions
Python is not my favourite language, but thewand binding for ImageMagick is really simple to use and to produce quickly good results.