Python script to pixelate images and video using K-Means clustering in the Lab colorspace. Video pixelating support multi-processing to achieve better performance.

#!/bin/sh

# Mandatory
sudo apt-get install python-numpy python-sklearn python-opencv

# Optional
sudo apt-get install libav-tools

• pypxl.py: Pixelate library
• pypxl_image.py: Pixelate an image from command line
• pypxl_video.py: Pixelate a video from command line

Main options are -s (number of pixels in the resulting image) and -k (number of clusters in K-Means). Usage example:

#!/bin/sh

# Pixelate and display an simage
./pypxl_image.py -k 6 -s 4 4 test.png

Video pixelate has the same options as images, plus some neat tricks.

K-Means is an expensive task, so you may consider to use multi-processing in pixelating videos (-np option). This may not be efficient enough: in that case you can apply a pre-scale to each frame to reduce the time spent on K-Means (-p option).

pypxl_video.py can process only video tracks. The audio can be extracted and re-added to the pixelated video with avconv:

#!/bin/sh

# Extract audio from source video
avconv -i src.avi -c:a copy src_audio.mp3

# Pixelate source video using 4 parallel processes
./pypxl_video.py -np 4 -s 8 8 src.avi out_noaudio.avi

# Add audio to pixelate video
avconv -i out_noaudio.avi -i src_audio.mp3 out.avi