def _inccs(self, image, ccs=None):
if ccs is None:
ccs = set(image.cc_analysis())
spikes = self._possible_text_areas(image=image, ccs=ccs)
inccs = ccs_in_rspike(spikes, ccs)
return inccs
def _inccs(self,image,ccs=None):
if ccs is None:
ccs = set(image.cc_analysis())
spikes = self._possible_text_areas(image=image,
ccs=ccs)
inccs = ccs_in_rspike(spikes,ccs)
return inccs
def _possible_text_areas(self, image=None, ccs=None):
image = self._def_image(image)
if ccs is None:
ccs = image.cc_analysis()
p, spikes = self._possible_text_areas_projection(image)
for s in spikes[:]:
cs = ccs_in_rspike([s], ccs)
avgaspect = median([c.aspect_ratio()[0] for c in cs])
if not (len(cs) > self._min_cc_count and between(
avgaspect, self._avg_cutoff[0], self._avg_cutoff[1])):
self.l.debug("Avg aspect: %f, %d ccs", avgaspect, len(cs))
spikes.remove(s)
return spikes
def _possible_text_areas(self,
image=None,
ccs=None):
image=self._def_image(image)
if ccs is None:
ccs = image.cc_analysis()
p,spikes = self._possible_text_areas_projection(image)
for s in spikes[:]:
cs = ccs_in_rspike([s],ccs)
avgaspect = median([ c.aspect_ratio()[0] for c in cs ])
if not (len(cs) > self._min_cc_count and
between(avgaspect,self._avg_cutoff[0],self._avg_cutoff[1])):
self.l.debug("Avg aspect: %f, %d ccs",avgaspect,len(cs))
spikes.remove(s)
return spikes
r =mi.with_row_projections(fac=fac)
r = mi.draw_y_proj(p,image=r,color=RGBPixel(100,0,0),fac=fac)
r.save_PNG("%s/%s-01-rowprojections.png"%(outputbase,basename))
logging.debug("01-rowprojections.png")
# draw in text areas based on these projections
r = mi.to_rgb()
width = mi._image.ncols-1
for re in p.rspikes(width):
r.draw_hollow_rect(re,RGBPixel(210,210,210))
for re in sp:
r.draw_hollow_rect(re,RGBPixel(150,150,150))
ra = t._possible_text_areas()
for re in ra:
r.draw_hollow_rect(re,RGBPixel(255,0,0))
firstccs = ccs_in_rspike(ra,mi.without_insidestaves_info().cc_analysis())
for c in firstccs:
r.highlight(c,RGBPixel(0,100,0))
r.save_PNG("%s/%s-02-areas-and-ccs.png"%(outputbase,basename))
logging.debug("02-areas-and-ccs.png")
# Draw words
r = mi.to_rgb()
bsimg = mi.without_insidestaves_info()
inccs = t._inccs(image=bsimg)
for c in inccs:
r.highlight(c,RGBPixel(0,200,0))
words = wordlist(t._words(image=bsimg,ccs=inccs),inccs,t._min_wordlength)
for c in words:
r = mi.with_row_projections(fac=fac)
r = mi.draw_y_proj(p, image=r, color=RGBPixel(100, 0, 0), fac=fac)
r.save_PNG("%s/%s-01-rowprojections.png" % (outputbase, basename))
logging.debug("01-rowprojections.png")
# draw in text areas based on these projections
r = mi.to_rgb()
width = mi._image.ncols - 1
for re in p.rspikes(width):
r.draw_hollow_rect(re, RGBPixel(210, 210, 210))
for re in sp:
r.draw_hollow_rect(re, RGBPixel(150, 150, 150))
ra = t._possible_text_areas()
for re in ra:
r.draw_hollow_rect(re, RGBPixel(255, 0, 0))
firstccs = ccs_in_rspike(ra, mi.without_insidestaves_info().cc_analysis())
for c in firstccs:
r.highlight(c, RGBPixel(0, 100, 0))
r.save_PNG("%s/%s-02-areas-and-ccs.png" % (outputbase, basename))
logging.debug("02-areas-and-ccs.png")
# Draw words
r = mi.to_rgb()
bsimg = mi.without_insidestaves_info()
inccs = t._inccs(image=bsimg)
for c in inccs:
r.highlight(c, RGBPixel(0, 200, 0))
words = wordlist(t._words(image=bsimg, ccs=inccs), inccs,
t._min_wordlength)