def check_glyph_accent(item,glyph):
"""Check two glyphs for beeing grouped to one single character. This function is for unit connected-components like i, j or colon.
Signature:
``check_glyph_accent(item,glyph)``
with
*item*:
Some connected-component.
*glyph*:
Some connected-component.
There is returned an array with two elements. The first element keeps a list of characters (images that has been united to a single image) and the second image is a list of characters which has to be removed as these have been united to a single character.
"""
remove = []
add = []
result = []
if(glyph.contains_x(item.ul_x) or glyph.contains_x(item.lr_x) or glyph.contains_x(item.center_x)): ##nebeinander?
if(not(item.contains_y(glyph.ul_y) or item.contains_y(glyph.lr_y) or item.contains_y(glyph.center_y))): ##nicht y-dimensions ueberschneident
remove.append(item)
remove.append(glyph)
new = union_images([item,glyph])
add.append(new)
result.append(add) #result[0] == ADD
result.append(remove) #result[1] == REMOVE
return result
def __bar_candidate_grouping(ungrouped_bars):
"""
Groups bar candidates and returns grouped bars
"""
grouped_bars = []
grouped_bars.append(image_utilities.union_images(ungrouped_bars))
return grouped_bars
def _vector_process_f_clef(self, glyph):
# use only right portion of f clefs when getting center_of_mass
g = None
for gl in glyph.splitx(0.5):
gl_coords = self._convert_bb_to_coords(self._get_glyph_bb(gl))
# no g
if not g:
g = gl
# left of g
elif gl.offset_x + gl.ncols < g.offset_x:
pass
# x-intersecting g
elif self._x_intersecting_coords(gl_coords, g.offset_x,
g.offset_x + g.ncols, 0):
g = union_images([g, gl])
# right of g
elif gl.offset_x >= g.offset_x + g.ncols:
g = gl
y_add = g.offset_y - glyph.offset_y
return g, y_add
def _remove_high_density_CCs(self, image, bg):
CCs = image.cc_analysis()
new_CCs = []
for g in CCs:
# throw away glyphs
if g.compactness()[0] >= self.density:
continue
else:
new_CCs.append(g)
combine = union_images([bg] + new_CCs)
return combine
def check_upper_neighbors(item,glyph,line):
"""Check for small signs grouped beside each other like quotation marks.
Signature:
``check_upper_neighbors(item,glyph,line)``
with
*item*:
Some connected-component.
*glyph*:
Some connected-component.
*line*:
The ``Textline`` Object which includes ``item`` and ``glyph``
Returns an array with two elements. The first element keeps a list of
characters (images that has been united to a single image) and the
second image is a list of characters which has to be removed as
these have been united to a single character.
"""
remove = []
add = []
result = []
minheight = min([item.nrows,glyph.nrows])
# glyphs must be small, of similar size and on the same height
if(not(glyph.lr_y >= line.center_y and glyph.lr_y-(glyph.nrows/3) <= line.lr_y)):
if (glyph.contains_y(item.center_y) and item.contains_y(glyph.center_y)):
minwidth = min([item.ncols,glyph.ncols])
distance = item.lr_x - glyph.lr_x
if(distance > 0 and distance <= minwidth*3):
remove.append(item)
remove.append(glyph)
new = union_images([item,glyph])
add.append(new)
result.append(add) #result[0] == ADD
result.append(remove) #result[1] == REMOVE
return result
def get_line_glyphs(image,textlines):
"""Splits image regions representing text lines into characters.
Signature:
``get_line_glyphs (image, segments)``
with
*image*:
The document image that is to be further segmentated. It must contin the
same underlying image data as the second argument *segments*
*segments*:
A list ``Cc`` data types, each of which represents a text line region.
The image views must correspond to *image*, i.e. each pixels has a value
that is the unique label of the text line it belongs to. This is the
interface used by the plugins in the \"PageSegmentation\" section of the
Gamera core.
The result is returned as a list of Textline_ objects.
.. _Textline: gamera.toolkits.ocr.classes.Textline.html
"""
i=0
show = []
lines = []
ret,sub_ccs = image.sub_cc_analysis(textlines)
for ccs in sub_ccs:
line_bbox = Rect(textlines[i])
i = i + 1
glyphs = ccs[:]
newlist = []
remove = []
add = []
result = []
glyphs.sort(lambda x,y: cmp(x.ul_x, y.ul_x))
for position, item in enumerate(glyphs):
if(True):
#if(not(glyph.lr_y >= line_bbox.center_y and glyph.lr_y-(glyph.nrows/3) <= line_bbox.lr_y)): ## is this part of glyph higher then line.center_y ?
left = position - 2
if(left < 0):
left = 0
right = position + 2
if(right > len(glyphs)):
right = len(glyphs)
checklist = glyphs[left:right]
for glyph in checklist:
if (item == glyph):
continue
result = check_upper_neighbors(glyph,item,line_bbox)
if(len(result[0]) > 0): #something has been joind...
joind_upper_connection = result[0][0] #joind glyph
add.append(joind_upper_connection)
remove.append(result[1][0]) #first part of joind one
remove.append(result[1][1]) #second part of joind one
for glyph2 in checklist: #maybe the upper joind glyphs fits to a glyph below...
if(glyphs == joind_upper_connection):
continue
if(joind_upper_connection.contains_x(glyph2.center_x)): #fits for example on ae, oe, ue in german alph
new = union_images([glyph2,joind_upper_connection])
add.append(new)
remove.append(glyph2)
add.remove(joind_upper_connection)
break
for elem in remove:
if (elem in checklist):
checklist.remove(elem)
for glyph in checklist:
if(item == glyph):
continue
result = check_glyph_accent(item,glyph)
if(len(result[0]) > 0): #something has been joind...
add.append(result[0][0]) #joind glyph
remove.append(result[1][0]) #first part of joind one
remove.append(result[1][1]) #second part of joind one
for elem in remove:
if(elem in glyphs):
glyphs.remove(elem)
for elem in add:
glyphs.append(elem)
new_line = Textline(line_bbox)
final = []
if(len(glyphs) > 0):
for glyph in glyphs:
final.append(glyph)
new_line.add_glyphs(final,False)
new_line.sort_glyphs() #reading order -- from left to right
lines.append(new_line)
for glyph in glyphs:
show.append(glyph)
return lines
image_bin.reset_onebit_image()
# find likely rotation angle and correct
angle, tmp = image_bin.rotation_angle_projections()
image_bin = image_bin.rotate(angle=angle)
for i in range(filter_runs):
image_bin.filter_short_runs(filter_runs_amt, 'black')
image_bin.filter_narrow_runs(filter_runs_amt, 'black')
return image_bin
if __name__ == '__main__':
manuscript = 'stgall390'
all_files = os.listdir('./png/')
fnames = [x for x in all_files if 'text.png' in x and manuscript in x]
for filename in fnames:
print('processing ' + filename + '...')
raw_image = gc.load_image('./png/' + filename)
image, eroded, angle = preproc.preprocess_images(raw_image,
despeckle_amt=20,
filter_runs=0)
line_strips, lines_peak_locs, proj = preproc.identify_text_lines(
image, eroded)
unioned_lines = union_images(line_strips)
unioned_lines.save_image('cleaned_' + filename)