capture class
[imago.git] / imago.py
index 6ab51a1..ff49826 100755 (executable)
--- a/imago.py
+++ b/imago.py
 #!/usr/bin/env python
-"""Usage: imago.py file"""
+
+"""Go image recognition"""
 
 import sys
+import os
 import math
-import Image, ImageDraw
+import argparse
+from operator import itemgetter
+
+try:
+    import Image, ImageDraw
+except ImportError, msg:
+    print >> sys.stderr, msg
+    sys.exit(1)
+
 import im_debug
-import filter
+import filters
 from hough import Hough
 
-class UsageError(Exception):
-    def __init__(self, msg):
-        self.msg = msg
+Saving_dir = ''
+Saving_num = 0
 
-def main(*argv):
+def main():
     """Main function of the program."""
     
-    show_all = False
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('file', metavar='file', nargs=1,
+                        help="image to analyse")
+    parser.add_argument('-w', type=int, default=640,
+                        help="scale image to the specified width before analysis")
+    parser.add_argument('-d', '--debug', dest='show_all', action='store_true',
+                        help="show every step of the computation")
+    parser.add_argument('-s', '--save', dest='do_something', action='store_const',
+                        const=image_save, default=im_debug.show,
+                        help="save images instead of displaying them")
+    parser.add_argument('-v', '--verbose', dest='verbose', action='store_true',
+                        help="report progress")
+    args = parser.parse_args()
+
+    show_all = args.show_all
+    do_something = args.do_something
+    verbose = args.verbose
 
     try:
-        if argv is ():
-            argv = sys.argv[1:]
-            if argv == []:
-                raise UsageError('Missing filename')
-        if "--help" in argv:
-            print __doc__
-            return 0
-        if "--debug" in argv:
-            show_all = True
-    except UsageError, err:
-        print >>sys.stderr, err.msg, "(\"imago.py --help\" for help)"
-        return 2
-
-    try:
-        image = Image.open(argv[0])
+        image = Image.open(args.file[0])
     except IOError, msg:
-        print >>sys.stderr, msg
+        print >> sys.stderr, msg
         return 1
+    if image.mode == 'P':
+        image = image.convert('RGB')
+    
+    if image.size[0] > args.w:
+        image = image.resize((args.w, int((float(args.w)/image.size[0]) *
+                              image.size[1])), Image.ANTIALIAS)
+    global Saving_dir
+    Saving_dir = "saved/" + args.file[0][:-4] + "_" + str(image.size[0]) + "/"
+    
+    if verbose:
+        print >> sys.stderr, "preprocessing"
+
     if show_all:
-       im_debug.show(image, "original image")
+        do_something(image, "original image")
 
     im_l = image.convert('L')
     if show_all:
-        im_debug.show(im_l, "ITU-R 601-2 luma transform")
+        do_something(im_l, "ITU-R 601-2 luma transform")
+
+    if verbose:
+        print >> sys.stderr, "edge detection"
 
-    im_edges = filter.edge_detection(im_l)
+    im_edges = filters.edge_detection(im_l, 2)
     if show_all:    
-        im_debug.show(im_edges, "edge detection")
+        do_something(im_edges, "edge detection")
 
-    im_h = filter.high_pass(im_edges, 100)
+    im_h = filters.high_pass(im_edges, 100)
     if show_all:
-        im_debug.show(im_h, "high pass filter")
+        do_something(im_h, "high pass filters")
     
+    if verbose:
+        print >> sys.stderr, "hough transform"
+
     hough1 = Hough(im_h.size)
     im_hough = hough1.transform(im_h)
     if show_all:
-        im_debug.show(im_hough, "hough transform")
+        do_something(im_hough, "hough transform")
 
-    im_h2 = filter.high_pass(im_hough, 120)
+    im_hough = filters.peaks(im_hough)
     if show_all:
-        im_debug.show(im_h2, "second high pass filter")
-
-    hough2 = Hough(im_h2.size)
-    im_hough2 = hough2.transform(im_h2)
+        do_something(im_hough, "peak extraction")
+               
+    im_h2 = filters.high_pass(im_hough, 120)
     if show_all:
-        im_debug.show(im_hough2, "second hough transform")
+        do_something(im_h2, "second high pass filters")
 
-    im_h3 = filter.high_pass(im_hough2, 120)
+    im_h2 = filters.components2(im_h2)
     if show_all:
-       im_debug.show(im_h3, "third high pass filter")
-     
-    lines = hough2.find_angle_distance(im_h3)
-
-    im_lines = Image.new('L', im_h2.size)
+        do_something(im_h2, "components centers")
 
-    draw = ImageDraw.Draw(im_lines)
+    if verbose:
+        print >> sys.stderr, "second hough transform"
 
-    for line in lines:
-       draw.line(line_from_angl_dist(line, im_h2.size), fill=255)
+    hough2 = Hough(im_h2.size) 
+    # im_hough might be used instead im_h2, but at the moment it brings a lot of
+    # noise to the second transform, which later confuses the center-finding
+    # mechanism (which is not very robust yet)
+    im_hough2 = hough2.transform(im_h2)
     if show_all:
-       im_debug.show(im_lines, "lines")
+        do_something(im_hough2, "second hough transform")
 
-    im_c = combine(im_h2, im_lines)
+    im_h3 = filters.high_pass(im_hough2, 120)
+    if show_all:
+        do_something(im_h3, "third high pass filter")
+     
+    im_h3 = filters.components(im_h3)
     if show_all:
-        im_debug.show(im_c, "first hough x lines")
+        do_something(im_h3, "half centers")
+
+    if verbose:
+        print >> sys.stderr, "finding the grid"
+
+    lines_m = hough2.all_lines_h(im_h3)
+    lines = []
+    im_c = im_h2.convert('RGB').convert('RGB', (1, 0.5, 0.5, 0))
+    draw_c = ImageDraw.Draw(im_c)
+
+    for line_l in lines_m:
+        im_line = Image.new('L', im_h2.size)
+        draw = ImageDraw.Draw(im_line)
+        line_points = set()
+        for line in line_l:
+            draw.line(line_from_angl_dist(line, im_h2.size), fill=255, width=7)
+            draw_c.line(line_from_angl_dist(line, im_c.size), fill=(70, 70, 70), width=7)
+            for p in combine(im_h2, im_line):
+                line_points.add(p)
+        for point in line_points:
+            draw_c.point(point, fill=(120, 255, 120))
+        lines.append(hough1.lines_from_list(line_points))
 
-    collapse(im_c)
     if show_all:
-        im_debug.show(im_c, "optimalised hough")
+        do_something(im_c, "hough x lines")
 
-    lines = hough1.all_lines(im_c)
-    draw = ImageDraw.Draw(image)
-    for line in lines:
-       draw.line(line_from_angl_dist(line, image.size), fill=(120, 255, 120))
+    image_g = image.copy()
+    draw = ImageDraw.Draw(image_g)
+    for line in [l for s in lines for l in s]:
+        draw.line(line_from_angl_dist(line, image.size), fill=(120, 255, 120))
+    if show_all:
+        do_something(image_g, "the grid")
+
+    intersections = intersections_from_angl_dist(lines, image.size)
+    image_g = image.copy()
+    draw = ImageDraw.Draw(image_g)
+    for line in intersections:
+        for (x, y) in line:
+            draw.point((x , y), fill=(120, 255, 120))
+    
+    for line in intersections:
+        print ' '.join([stone_color(image, intersection) for intersection in
+                       line])
 
-    im_debug.show(image, "the grid")
+    if show_all:
+        do_something(image_g, "intersections")
 
     return 0
 
-def collapse(image):
-    #HACK
-    im_l = image.load()
-    last = False
-    for y in xrange(image.size[1]):
-       for x in xrange(image.size[0]):
-           if im_l[x,y] and last:
-                im_l[x, y] = 0
-               last = False
-           elif im_l[x, y]:
-               last = True
-           elif last:
-               last = False
+def stone_color(image, (x, y)):
+    suma = 0.
+    for i in range(-2, 3):
+        for j in range(-2, 3):
+            try:
+                suma += sum(image.getpixel((x + i, y + j)))
+            except IndexError:
+                pass
+    suma /= 3 * 25
+    if suma < 55:
+        return 'B'
+    elif suma < 200: 
+        return '.'
+    else:
+        return 'W'
+
+def image_save(image, title=''):
+    global Saving_dir
+    global Saving_num
+    filename = Saving_dir + "{0:0>2}".format(Saving_num) + '.jpg'
+    if not os.path.isdir(Saving_dir):
+        os.makedirs(Saving_dir)
+    image.save(filename, 'JPEG')
+    Saving_num += 1
 
 def combine(image1, image2):
     im_l1 = image1.load()
     im_l2 = image2.load()
 
-    im_n = Image.new('L', image1.size)
-    im_nl = im_n.load()
+    on_both = []
 
     for x in xrange(image1.size[0]):
         for y in xrange(image1.size[1]):
-           if im_l1[x, y] and im_l2[x, y]:
-               im_nl[x, y] = 255
-    return im_n
+            if im_l1[x, y] and im_l2[x, y]:
+                on_both.append((x, y))
+    return on_both
 
 def line_from_angl_dist((angle, distance), size):
     x1 = - size[0] / 2
-    y1 = int(round((x1 * math.sin(angle) - distance)/math.cos(angle))) + size[1] / 2
+    y1 = int(round((x1 * math.sin(angle) - distance) / math.cos(angle))) + size[1] / 2
     x2 = size[0] / 2 
-    y2 = int(round((x2 * math.sin(angle) - distance)/math.cos(angle))) + size[1] / 2
+    y2 = int(round((x2 * math.sin(angle) - distance) / math.cos(angle))) + size[1] / 2
     return [(0, y1), (size[0] - 1, y2)]
 
+def intersections_from_angl_dist(lines, size):
+    intersections = []
+    for (angl1, dist1) in sorted(lines[1], key=itemgetter(1)):
+        line = []
+        for (angl2, dist2) in sorted(lines[0], key=itemgetter(1)):
+            if abs(angl1 - angl2) > 0.4:
+                x =  - ((dist2 / math.cos(angl2))-(dist1 / math.cos(angl1))) / (math.tan(angl1) - math.tan(angl2))
+                y = (math.tan(angl1) * x) - (dist1 / math.cos(angl1))
+                if (-size[0] / 2 < x < size[0] / 2 and 
+                    -size[1] / 2 < y < size[1] / 2):
+                    line.append((int(x + size[0] / 2), int(y + size[1] / 2)))
+        intersections.append(line)
+    return intersections
+
 if __name__ == '__main__':
     sys.exit(main())