def calculate_single_weight(self, angle):
"""Finds the strongest gradient in the image"""
obj = ShearTool(self.img)
corrected = obj.shear(angle)
p = BinaryProjection(corrected, Projection.TYPE_HORIZONTAL)
# p.debug()
return max(np.gradient(p.get_projection())) + abs(min(np.gradient(p.get_projection())))
def calculate_single_weight(self, angle):
"""Finds the strongest gradient in the image"""
obj = ShearTool(self.img)
corrected = obj.shear(angle)
p = BinaryProjection(corrected, Projection.TYPE_HORIZONTAL)
# p.debug()
return max(np.gradient(p.get_projection())) + abs(min(np.gradient(p.get_projection())))
class AreaFactory:
"""
>>> arr = cv2.imread('assets/img/doctests/digits.jpg', 0)
>>> f = AreaFactory(arr, Projection.TYPE_VERTICAL, Rectangle.from_ndarray(arr))
>>> list(f.find_areas())
[<Rectangle 0+572, 0+1310>]
>>> # f.debug('area factory vertical')
>>>
>>> arr = cv2.imread('assets/img/doctests/digits.jpg', 0)
>>> f = AreaFactory(arr, Projection.TYPE_HORIZONTAL, Rectangle.from_ndarray(arr))
>>> list(f.find_areas())
[<Rectangle 0+572, 184+838>]
>>> # f.debug('area factory horizontal')
"""
img = None
projection = None
basis = None
measurement = -1
def __init__(self, full_color_image, orientation, basis):
# Projection(full_color_image, orientation, 0.2).debug()
self.img = full_color_image
self.projection = BinaryProjection(full_color_image, orientation)
self.basis = basis or Rectangle.from_ndarray(full_color_image)
self.measurement = full_color_image.shape[orientation]
def find_areas(self, overflow_percent=0.03):
"""
:param overflow_percent: 0 to 1
:return: [(10, 100), (150, 300)]
"""
ranges = self._split_peaks_into_ranges()
return self._convert_ranges_into_coordinates(self.basis, ranges)
def get_projection(self):
return self.projection.get_projection()
def _split_peaks_into_ranges(self):
peaks = list(self.projection.get_peaks())
if len(peaks) % 2:
peaks = peaks[1:]
if len(peaks) < 2:
yield 0, self.projection.measurement
else:
for i in xrange(0, len(peaks), 2):
yield peaks[i][0], peaks[i+1][0]
def _convert_ranges_into_coordinates(self, basis, ranges):
for first_scaled, second_scaled in ranges:
if self.projection.orientation == Projection.TYPE_VERTICAL:
yield Rectangle(basis.top + first_scaled, second_scaled - first_scaled,
basis.left, basis.width)
else:
yield Rectangle(basis.top, basis.height,
basis.left + first_scaled, second_scaled - first_scaled)
def debug(self, window_title="Area Factory"):
areas = list(self.find_areas()) # todo why casting to list is required?
Plotter.outline_rectangles(self.img, areas, window_title=window_title)
def __init__(self, full_color_image, orientation, basis):
# Projection(full_color_image, orientation, 0.2).debug()
self.img = full_color_image
self.projection = BinaryProjection(full_color_image, orientation)
self.basis = basis or Rectangle.from_ndarray(full_color_image)
self.measurement = full_color_image.shape[orientation]
class AreaFactory:
"""
>>> arr = cv2.imread('assets/img/doctests/digits.jpg', 0)
>>> f = AreaFactory(arr, Projection.TYPE_VERTICAL, Rectangle.from_ndarray(arr))
>>> list(f.find_areas())
[<Rectangle 0+572, 0+1310>]
>>> # f.debug('area factory vertical')
>>>
>>> arr = cv2.imread('assets/img/doctests/digits.jpg', 0)
>>> f = AreaFactory(arr, Projection.TYPE_HORIZONTAL, Rectangle.from_ndarray(arr))
>>> list(f.find_areas())
[<Rectangle 0+572, 184+838>]
>>> # f.debug('area factory horizontal')
"""
img = None
projection = None
basis = None
measurement = -1
def __init__(self, full_color_image, orientation, basis):
# Projection(full_color_image, orientation, 0.2).debug()
self.img = full_color_image
self.projection = BinaryProjection(full_color_image, orientation)
self.basis = basis or Rectangle.from_ndarray(full_color_image)
self.measurement = full_color_image.shape[orientation]
def find_areas(self, overflow_percent=0.03):
"""
:param overflow_percent: 0 to 1
:return: [(10, 100), (150, 300)]
"""
ranges = self._split_peaks_into_ranges()
return self._convert_ranges_into_coordinates(self.basis, ranges)
def get_projection(self):
return self.projection.get_projection()
def _split_peaks_into_ranges(self):
peaks = list(self.projection.get_peaks())
if len(peaks) % 2:
peaks = peaks[1:]
if len(peaks) < 2:
yield 0, self.projection.measurement
else:
for i in xrange(0, len(peaks), 2):
yield peaks[i][0], peaks[i + 1][0]
def _convert_ranges_into_coordinates(self, basis, ranges):
for first_scaled, second_scaled in ranges:
if self.projection.orientation == Projection.TYPE_VERTICAL:
yield Rectangle(basis.top + first_scaled,
second_scaled - first_scaled, basis.left,
basis.width)
else:
yield Rectangle(basis.top, basis.height,
basis.left + first_scaled,
second_scaled - first_scaled)
def debug(self, window_title="Area Factory"):
areas = list(
self.find_areas()) # todo why casting to list is required?
Plotter.outline_rectangles(self.img, areas, window_title=window_title)
def __init__(self, full_color_image, orientation, basis):
# Projection(full_color_image, orientation, 0.2).debug()
self.img = full_color_image
self.projection = BinaryProjection(full_color_image, orientation)
self.basis = basis or Rectangle.from_ndarray(full_color_image)
self.measurement = full_color_image.shape[orientation]