hocr-tools extension finding the print space of a page w/o margin notes

hocr-printspace

#!/usr/bin/env python

# Find the print space of a page, without margin notes

import sys,os,string,re
from lxml import html
import argparse
from PIL import Image, ImageDraw

def get_prop(node,name):
    title = node.get('title')
    props = title.split(';')
    for prop in props:
        (key,args) = prop.split(None,1)
        if key==name: return args

def get_text(node):
    textnodes = node.xpath(".//text()")
    s = string.join([text for text in textnodes])
    return re.sub(r'\s+',' ',s)

def get_bbox(node):
    bbox = get_prop(node,'bbox')
    if not bbox: return None
    return tuple([int(x) for x in bbox.split()])

def guess_lower_boundary(freq, min):
    candidates = sorted(freq, key=freq.get, reverse=True)
    return [v for v in candidates if v >= min][0]

parser = argparse.ArgumentParser(description="Find print space of a page")
parser.add_argument("-n","--normalize_amount",type=int,default=10,
                    help="Amount to normalize coordinates with, default: %(default)s")
parser.add_argument("-D","--doublepage",default=False,action="store_true",
                    help="Whether pages are single or double, default: %(default)s")
parser.add_argument("-i","--inverse_vertically",default=False,action="store_true",
                    help="Change to measure the vertical coordinate from the bottom instead of the top")

parser.add_argument("hocrfile", help="hOCR file to detect print space in")
parser.add_argument("imagefile", help="corresponding image file")

args = parser.parse_args()
NORMALIZE_AMOUNT = args.normalize_amount

stream = open(args.hocrfile)
doc = html.fromstring(stream.read())
im = Image.open(args.imagefile)
dr = ImageDraw.Draw(im)
width, height = im.size

pages = doc.xpath("//*[@class='ocr_page']")

for page in pages:
    print('Page "%s"'%page.get('title'))
    lines = page.xpath("//*[@class='ocr_line']")
    coord_freq = [{}, {}, {}, {}]
    coord_list = []
    length_freq = {}
    empty_lines = 0
    marginal_lines = 0
    for line in lines:
        if not get_text(line).strip():
            empty_lines += 1
            continue
        coords = get_bbox(line)
        if args.inverse_vertically:
            coords = list(coords)
            coords[1] = height-coords[1]
            coords[3] = height-coords[3]
            coords = tuple(coords)
        # discard small textblock on the margins
        if coords[2]<100:
            #print coords
            marginal_lines += 1
            continue
        if coords[0]>width-100:
            #print coords
            marginal_lines += 1
            continue
        #print "%s"%get_text(line)
        coord_list.append(coords)
        #print coords
        length = ( int(coords[2]) - int(coords[0]) ) / NORMALIZE_AMOUNT * NORMALIZE_AMOUNT
        #print length
#        dr.rectangle(coords, fill=None, outline = "blue")
        if not length in length_freq:
            length_freq[length] = 0
        length_freq[length] += 1
        for pos in range(0,4):
            if not coords[pos] in coord_freq[pos]:
                coord_freq[pos][coords[pos]] = 0
            coord_freq[pos][coords[pos]] += 1
    common_length = sorted(length_freq, key=length_freq.get, reverse=True)[0]

    first = False
    last = False
    for coords in coord_list:
        w = ( int(coords[2]) - int(coords[0]) ) / NORMALIZE_AMOUNT * NORMALIZE_AMOUNT
        if w == common_length:
            dr.rectangle(coords, fill=None, outline = "blue")
            if not first:
                print "defifnining first"
                first = (coords[2], coords[1])
                print first
            else:
                last = (coords[2], coords[1])
        else:
            dr.rectangle(coords, fill=None, outline = "red")
    print first
    print last
    if (first[0] == last[0]):
        dr.line([first[0], 0, first[0], height], fill=0, width=3)

    else:
        print "else"
        a = (first[1]-last[1])/(first[0]-last[0])
        b = first[1]-first[0]*a
        zero = -b/a
        xheight = (height-b)/a
        dr.line([zero, 0, xheight, height], fill=0, width=3)

    if False:
        x0 = guess_lower_boundary(coord_freq[0], 0)
        y0 = sorted(coord_freq[1], key=coord_freq[1].get)[0]
        x1 = guess_lower_boundary(coord_freq[2], 0)
        y1 = sorted(coord_freq[3], key=coord_freq[3].get, reverse=True)[0]
        dr.rectangle([x0, y0, x1, y1], fill=None, outline="red")


print "%s empty lines skipped"%empty_lines
print "%s marginal lines skipped"%marginal_lines
#print length_freq
#print "\t[%s, %s]"%(x0, y0)
#print "\t[%s, %s]"%(x1, y1)
print common_length

im.save("rectangle.png")

Here are three examples with bounding boxes: