def _text_recognition_dump(result, args=None, kwargs=None, dump_dir=None):
cbounds = Config.instance().config('argv:cbounds', (0, 0, 255)).value
image = args[1]
storage = kwargs['storage']['TextRecognitor']
text, text_candidates = storage['text'], storage['text_candidates']
graphical = storage['graphical']
# Dump text region candidates
image1 = image.copy()
for t in text_candidates:
t.display_bounds(image1, color=cbounds)
image1.save(os.path.join(dump_dir, 'candidates.png'))
# Dump recognized text regions
image2 = image.copy()
for t in text:
t.display_bounds(image2, color=cbounds)
image2.save(os.path.join(dump_dir, 'recognized.png'))
# Dump difference text candidates and recognized text regions
ImageChops.difference(image1, image2).save(
os.path.join(dump_dir, 'difference.png'))
# Dump remaining graphical segments
image3 = Image.create(image.mode, image.width, image.height, background=(255, 255, 255))
for g in graphical:
g.display(image3, color=g.color)
image3.save(os.path.join(dump_dir, 'graphical.png'))
def toimage(self, mode="L", color=255, background=0, border=0, angle=None):
"""Convert this segment to image.
:param mode: mode of resulting image
:param color: color of segment pixels on resulting image
:param border: image border width (segment pixels will be surrounded by
border of background color if value is greater than 0)
:param angle: can be used to create rotated image (usefull for OCR to
recognize vertical text segments by rotating them to be a
horizontal text segments)"""
result = Image.create(mode, self.width + 2 * border, self.height + 2 * border, background=background)
p = result.pixels
l, t = self.left, self.top
for x, y in self.area:
p[x - l + border, y - t + border] = color
if angle:
return result.rotate(angle)
else:
return result
def run(self):
"""Execute application and return exit code."""
# Load source image
source = Image.load(self.infile)
if source.mode != 'RGB':
source = source.convert('RGB')
# Create filter stack
f = None
f = Parser(next_filter=f)
f = FigureRecognitor(next_filter=f)
f = TextRecognitor(next_filter=f)
f = Segmentizer(next_filter=f)
f = Quantizer(next_filter=f)
# Execute filter stack
f(source, storage={}, key=source.checksum()).save(self.outfile)
return 0
def __create_result_image(self, image, clusters):
"""Create result image by changing color of each pixel in source image
to one of cluster centroid colors."""
encoder = self.__c_encoder
decoder = self.__c_decoder
res = Image.create(image.mode, image.width, image.height)
dpix = res.pixels
spix = image.pixels
cache = {}
for y in xrange(image.height):
for x in xrange(image.width):
s = spix[x, y]
if s not in cache:
for c in clusters:
if encoder(s) in c.samples:
val = decoder(c.centroid)
break
cache[s] = val
dpix[x, y] = val
else:
dpix[x, y] = cache[s]
return res