get_images.py

# 'doc' is ALTO XML, parsed into an etree xml obj.

def get_illustration_coords(doc, component="PrintSpace"):
  page = doc.find("Layout/Page")
  illustrations = doc.findall('Layout/Page/{0}/ComposedBlock[@TYPE="Illustration"]/GraphicalElement'.format(component))
  pageh, pagew = int(page.attrib['HEIGHT']), int(page.attrib['WIDTH'])
  images = []
  for img in illustrations:
    x,y = map(int, [img.attrib['HPOS'], img.attrib['VPOS']])
    h,w = map(int, [img.attrib['HEIGHT'], img.attrib['WIDTH']])
    images.append([x,y,w,h])
  return (pagew, pageh), images