def find_columns(self):
"""Get columns in a section of the image
"""
portion = iulib.bytearray()
iulib.extract_subimage(portion, self.inverted, 0, 0,
self.inverted.dim(0), self.topptr)
projection = high_pass_median(iulib.numpy(portion).sum(axis=1), 0.20)
posscols = self.get_possible_columns(projection)
bestcols = self.filter_columns(posscols, int(self._params.get("columns", 1)))
self.columns.extend(bestcols)
def find_columns(self):
"""Get columns in a section of the image
"""
portion = iulib.bytearray()
iulib.extract_subimage(portion, self.inverted, 0, 0,
self.inverted.dim(0), self.topptr)
projection = high_pass_median(iulib.numpy(portion).sum(axis=1), 0.20)
posscols = self.get_possible_columns(projection)
bestcols = self.filter_columns(posscols,
int(self._params.get("columns", 1)))
self.columns.extend(bestcols)
def init(self):
"""Initialise on receipt of the input."""
# pointer to the region that remains
# to be segmented - starts at the top
self.topptr = self.inarray.dim(1)
# obtain an inverted version of the array
self.inverted = iulib.bytearray()
self.inverted.copy(self.inarray)
iulib.binary_invert(self.inverted)
self.calc_bounding_boxes()
# list of extracted line rectangles
self.textlines = []
self.columns = []
def init(self):
"""Initialise on receipt of the input."""
# pointer to the region that remains
# to be segmented - starts at the top
self.topptr = self.inarray.dim(1)
# obtain an inverted version of the array
self.inverted = iulib.bytearray()
self.inverted.copy(self.inarray)
iulib.binary_invert(self.inverted)
self.calc_bounding_boxes()
# list of extracted line rectangles
self.textlines = []
self.columns = []
def remove_border(narray, average_char_height):
"""Try and remove anything that's in a likely
border region and return the subimage."""
na = iulib.numpy(narray)
hpr = na.sum(axis=0)
vpr = na.sum(axis=1)
hhp = high_pass_median(hpr, 5.0 / average_char_height)
vhp = high_pass_median(vpr, 5.0 / average_char_height)
vidx = vhp.nonzero()[0]
hidx = hhp.nonzero()[0]
b = iulib.bytearray()
iulib.extract_subimage(b, narray, int(vidx[0]), int(hidx[0]),
int(vidx[-1]), int(hidx[-1]))
return b
def remove_border(narray, average_char_height):
"""Try and remove anything that's in a likely
border region and return the subimage."""
na = iulib.numpy(narray)
hpr = na.sum(axis=0)
vpr = na.sum(axis=1)
hhp = high_pass_median(hpr, 5.0 / average_char_height)
vhp = high_pass_median(vpr, 5.0 / average_char_height)
vidx = vhp.nonzero()[0]
hidx = hhp.nonzero()[0]
b = iulib.bytearray()
iulib.extract_subimage(b, narray, int(vidx[0]), int(hidx[0]),
int(vidx[-1]), int(hidx[-1]))
return b
def find_lines(self):
"""
Get lines in a section of the images.
"""
for colrect in self.columns:
newrect = Rectangle(colrect.x0, 0, colrect.x1, self.topptr)
if newrect.area() < 1:
continue
portion = iulib.bytearray()
iulib.extract_subimage(portion, self.inverted, *newrect.points())
regions = get_lines_by_projection(
portion, float(self._params.get("highpass")))
plines = []
for bottom, top in regions:
height = top - bottom
if height - self.avgheight < self.avgheight / 3:
continue
plines.append(Rectangle(colrect.x0, bottom, colrect.x1, top))
cpline = None
clline = Rectangle(0, 0, 0, 0)
charboxes = self.get_char_boxes(self.boxes)
colboxes = [b for b in charboxes \
if b.overlaps(colrect.grow(10, 10))]
colboxes.sort(lambda x, y: cmp(x.y1, y.y1))
colboxes.reverse()
clines = []
for p in plines:
clines.append(Rectangle(0, 0, 0, 0))
while colboxes:
char = colboxes.pop(0)
cline = Rectangle(0, 0, 0, 0)
for i in range(len(plines)):
pline = plines[i]
if char.overlaps(pline):
clines[i].include(char)
self.textlines.extend(clines)
def find_lines(self):
"""
Get lines in a section of the images.
"""
for colrect in self.columns:
newrect = Rectangle(colrect.x0, 0, colrect.x1, self.topptr)
if newrect.area() < 1:
continue
portion = iulib.bytearray()
iulib.extract_subimage(portion, self.inverted, *newrect.points())
regions = get_lines_by_projection(portion, float(self._params.get("highpass")))
plines = []
for bottom, top in regions:
height = top - bottom
if height - self.avgheight < self.avgheight / 3:
continue
plines.append(Rectangle(colrect.x0, bottom, colrect.x1, top))
cpline = None
clline = Rectangle(0, 0, 0, 0)
charboxes = self.get_char_boxes(self.boxes)
colboxes = [b for b in charboxes \
if b.overlaps(colrect.grow(10, 10))]
colboxes.sort(lambda x, y: cmp(x.y1, y.y1))
colboxes.reverse()
clines = []
for p in plines:
clines.append(Rectangle(0, 0, 0, 0))
while colboxes:
char = colboxes.pop(0)
cline = Rectangle(0, 0, 0, 0)
for i in range(len(plines)):
pline = plines[i]
if char.overlaps(pline):
clines[i].include(char)
self.textlines.extend(clines)
