Markov image generator

Main usage: building Markov models from images, and generating random images.

Markov chain Python class:

• MarkovImg(order)

Methods:

• learn(img, salt) creates a model given an image img and salt value.

• generate(img_size) generates a random image with a size img_size.

The order>=1 parameter corresponds to the Markov chain order.

The 0<=salt<=1 parameter helps with learning by randomly flipping the LSB in pixels.

Different values of the parameters will give different results.

Basic usage

from markovimg import MarkovImg
from PIL import Image

img = Image.open("./test_images/lena.png")
img.show()
print(img.format, img.size, img.mode)
order,salt = 4,0.1
mc = MarkovImg(order)
mc.learn(img,salt)
im = mc.generate(img.size)
im.show()

This example uses the standard lena.png with order = 4, salt = 0.1:

to produce a random image like:

One can also include successive image transformations as following:

from markovimg import MarkovImg
from PIL import Image

fname = "./test_images/mandrill.png"
img = Image.open(fname)
img.show()
order,salt = 3,0.4
mc = MarkovImg(order)
mc.learn(img,salt)
mc.learn(img.rotate(90),salt)
mc.learn(img.rotate(180),salt)
mc.learn(img.rotate(270),salt)
mc.learn(img.transpose(Image.FLIP_LEFT_RIGHT),salt)
mc.learn(img.transpose(Image.FLIP_TOP_BOTTOM),salt)
mc.learn(img.transpose(Image.ROTATE_90),salt)
mc.learn(img.transpose(Image.ROTATE_180),salt)
mc.learn(img.transpose(Image.ROTATE_270),salt)
im = mc.generate((img.size[0]*2,2*img.size[1]))
im.show()

This example uses the standard mandrill.png with order=3, salt = 0.4:

to produce a random image like: