hough.py

   1 from math import sin, cos, pi
   2
   3 from PIL import Image
   4
   5 from commons import clear
   6
   7 class Hough:
   8     def __init__(self, size):
   9         self.size = size
  10         self.dt = pi / size[1]
  11         self.initial_angle = (pi / 4) + (self.dt / 2)
  12
  13     def transform(self, image):
  14         image_l = image.load()
  15         size = image.size
  16
  17         matrix = [[0]*size[1] for _ in xrange(size[0])]
  18
  19         dt = self.dt
  20         initial_angle = self.initial_angle
  21
  22         for x in xrange(size[0]):
  23             clear()
  24             print "hough transform: {0:>3}/{1}".format(x + 1, size[0])
  25             for y in xrange(size[1]):
  26                 if image_l[x, y]:
  27                     # for every angle:
  28                     for a in xrange(size[1]):
  29                         # find the distance:
  30                         # distance is the dot product of vector (x, y) - centerpoint
  31                         # and a unit vector orthogonal to the angle
  32                         distance = (((x - (size[0] / 2)) * sin((dt * a) + initial_angle)) +
  33                                     ((y - (size[1] / 2)) * -cos((dt * a) + initial_angle)) +
  34                                     size[0] / 2)
  35                         # column of the matrix closest to the distance
  36                         column = int(round(distance))
  37                         if column >= 0 and column < size[0]:
  38                             matrix[column][a] += 1
  39
  40         new_image = Image.new('L', size)
  41         new_image_l = new_image.load()
  42
  43         minimum = min([min(m) for m in matrix])
  44
  45         maximum = max([max(m) for m in matrix]) - minimum
  46
  47         for y in xrange(size[1]):
  48             for x in xrange(size[0]):
  49                 new_image_l[x, y] = (float(matrix[x][y] - minimum) / maximum) * 255
  50
  51         return new_image
  52
  53     def all_lines(self, image):
  54         im_l = image.load()
  55         lines = []
  56         for x in xrange(image.size[0]):
  57             for y in xrange(image.size[1]):
  58                 if im_l[x, y]:
  59                     lines.append(self.angle_distance((x, y)))
  60         return lines
  61
  62     def find_angle_distance(self, image):
  63         image_l = image.load()
  64
  65         points = []
  66
  67         count = 0
  68         point_x = 0
  69         point_y = 0
  70         for x in xrange(image.size[0] / 2):
  71             for y in xrange(image.size[1] / 2, image.size[1]):
  72                 if image_l[x, y]:
  73                     count += 1
  74                     point_x += x
  75                     point_y += y
  76         points.append((float(point_x) / count, float(point_y) / count))
  77
  78         count = 0
  79         point_x = 0
  80         point_y = 0
  81         for x in xrange(image.size[0] / 2, image.size[0]):
  82             for y in xrange(image.size[1] / 2, image.size[1]):
  83                 if image_l[x, y]:
  84                     count += 1
  85                     point_x += x
  86                     point_y += y
  87         points.append((float(point_x) / count, float(point_y) / count))
  88
  89         return [self.angle_distance(p) for p in points]
  90
  91     def angle_distance(self, point):
  92         return (self.dt * point[1] + self.initial_angle, point[0] - self.size[0] / 2)
  93