# A fragment of a Python script that can act as a front end for
# the assignment code.  This code is intended to be used in a 
# command-line-driven implementation of the algorithm.

import Image
import PIL_usm

# Load in the C++ support module

import painthelp

# The main painting algorithm described in the paper.  'src' is an
# image loaded using PIL, and Rs is a Python list of brush radii.

def paint( src, Rs ):
	# Construct a new image to serve as the canvas.  For every
	# brush size, we'll pass in the current canvas and get back
	# a canvas with new strokes painted onto it.

	canvas = Image.new( 'RGB', src.size )

	# Note that Hertzmann mentions a special colour that's maximally
	# far away from every other colour.  You'll need to figure out 
	# some way to make that work.

	for R in Rs:
		# Construct a Gaussian blurred version of the source image.
		ref = PIL_usm.gblur( src, f_sigma * R )

		w,h = canvas.size
		# Turn the image data in canvas into a Python string.
		# As long as the image is in RGB format, you'll be able
		# to read this string in C++ as a sequence of RGB byte 
		# triples.
		cts = canvas.tostring()

		# Call the paintLayer function in the C++ code and turn 
		# the result into a new PIL image.
		canvas = Image.fromstring( 'RGB', (w,h), 
			painthelp.paintLayer( cts, ref.tostring(), w, h, R ) )

	return canvas