def process(self):
"""
Segment with ocropy
"""
for (n, input_file) in enumerate(self.input_files):
log.info("INPUT FILE %i / %s", n, input_file)
downloaded_file = self.workspace.download_file(input_file)
log.info("downloaded_file %s", downloaded_file)
pcgts = page_from_file(downloaded_file)
page_width = pcgts.get_Page().get_imageWidth()
page_height = pcgts.get_Page().get_imageHeight()
# TODO binarized variant from get_AlternativeImage()
image_url = pcgts.get_Page().imageFilename
log.info("pcgts %s", pcgts)
binary = ocrolib.read_image_binary(
self.workspace.download_url(image_url))
binary = 1 - binary
scale = self.parameter['scale'] if self.parameter[
'scale'] != 0 else psegutils.estimate_scale(binary)
log.debug(binary)
pseg = self.compute_segmentation(binary, scale)
log.debug("pseg=%s", pseg)
# TODO reading order / enumber
# log.debug("finding reading order")
# lines = psegutils.compute_lines(pseg, scale)
# order = psegutils.reading_order([l.bounds for l in lines])
# lsort = psegutils.topsort(order)
regions = ocrolib.RegionExtractor()
regions.setPageLines(pseg)
dummyRegion = TextRegionType(
id="dummy",
Coords=CoordsType(
points="0,0 %s,0 %s,%s 0,%s" %
(page_width, page_width, page_height, page_height)))
pcgts.get_Page().add_TextRegion(dummyRegion)
for lineno in range(1, regions.length()):
log.debug("id=%s bbox=%s", regions.id(lineno),
regions.bbox(lineno))
textline = TextLineType(
id=concat_padded("line", lineno),
Coords=CoordsType(
points=points_from_y0x0y1x1(regions.bbox(lineno))))
dummyRegion.add_TextLine(textline)
ID = concat_padded(self.output_file_grp, n)
self.workspace.add_file(ID=ID,
file_grp=self.output_file_grp,
mimetype=MIMETYPE_PAGE,
local_filename="%s/%s.xml" %
(self.output_file_grp, ID),
content=to_xml(pcgts))
class OcropusSegmentPageBase(OcropusBase, base.JSONWriterMixin):
"""
Segment an image using Ocropus.
"""
abstract = True
stage = stages.PAGE_SEGMENT
intypes = [ocrolib.numpy.ndarray]
outtype = dict
def null_data(self):
"""
Return an empty list when ignored.
"""
return dict(columns=[], lines=[], paragraphs=[])
def process(self, input):
"""
Segment a binary image.
input: a binary image.
return: a dictionary of box types:
lines
paragraphs
columns
images
"""
out = dict(bbox=[0, 0, input.shape[1], input.shape[0]],
columns=[],
lines=[],
paragraphs=[])
try:
page_seg = self._comp.segment(input)
except (IndexError, TypeError, ValueError), err:
raise OcropusNodeError(err.message, self)
regions = ocrolib.RegionExtractor()
exfuncs = dict(lines=regions.setPageLines,
paragraphs=regions.setPageParagraphs)
# NB: These coordinates are relative to the TOP of the page
# for some reason
for box, func in exfuncs.iteritems():
func(page_seg)
for i in range(1, regions.length()):
out[box].append((regions.x0(i), regions.y0(i), regions.x1(i),
regions.y1(i)))
return out
def __init__(self, *args, **kwargs):
super(SegmentPageManual, self).__init__(*args, **kwargs)
self._regions = ocrolib.RegionExtractor()
self._segmenter = ocrolib.SegmentPageByRAST1()
def processPngFile(outRoot, origFile, fileNum):
baseName = os.path.basename(origFile)
baseBase, _ = os.path.splitext(baseName)
outDir = os.path.join(outRoot, "%s.%03d" % (baseBase, fileNum))
inFile = os.path.join(outDir, baseName)
os.makedirs(outDir, exist_ok=True)
shutil.copy(origFile, inFile)
inBase, _ = ocrolib.allsplitext(inFile)
print("** inBase=%s" % inBase)
# print("** binBase=%s" % binBase)
fname = inFile
outputdir = inBase
binFile = inBase + ".bin.png"
outFile = inBase + ".out.png"
outRoot2, outDir2 = os.path.split(outRoot)
outFile2 = os.path.join(outRoot2, "%s.out" % outDir2, baseName)
print("outFile2=%s" % outFile2)
# assert False
grayFile = inBase + ".nrm.png"
psegFile = inBase + ".pseg.png"
print(" inFile=%s" % inFile)
print(" binFile=%s" % binFile)
print("grayFile=%s" % grayFile)
print(" outFile=%s" % outFile)
assert inFile and binFile
assert outFile != inFile
assert outFile != binFile
if not binarize(inFile, binFile, grayFile):
binExists = os.path.exists(binFile)
print("Couldn't binarize inFile=%s binFile=%s exists=%s" %
(inFile, binFile, binExists))
return False
binary = ocrolib.read_image_binary(binFile)
print("$$ %s=%s" % (binFile, desc(binary)))
height, width = binary.shape
checktype(binary, ABINARY2)
check = check_page(np.amax(binary) - binary)
if check is not None:
print("%s SKIPPED %s (use -n to disable this check)" % (inFile, check))
return False
# if args.gray:
# if os.path.exists(base+".nrm.png"):
# gray = ocrolib.read_image_gray(base+".nrm.png")
# checktype(gray, GRAYSCALE)
# else:
# print_error("Grayscale version %s.nrm.png not found. Use ocropus-nlbin for creating " +
# "normalized grayscale version of the pages as well." % base)
# return
binary = 1 - binary # invert
scale = psegutils.estimate_scale(binary)
print("scale %f" % scale)
if np.isnan(scale) or scale > 1000.0:
print("%s: bad scale (%g); skipping\n" % (fname, scale))
return False
# find columns and text lines
print("computing segmentation")
segmentation = compute_segmentation(binary, scale)
if np.amax(segmentation) > maxlines:
print("%s: too many lines %g" % (fname, np.amax(segmentation)))
return False
print("segmentation=%s" % desc(segmentation))
print("number of lines %g" % np.amax(segmentation))
# compute the reading order
print("finding reading order")
lines = psegutils.compute_lines(segmentation, scale)
order = psegutils.reading_order([l.bounds for l in lines])
lsort = psegutils.topsort(order)
print("$$ lsort = %d = %s...%s" % (len(lsort), lsort[:10], lsort[-10:]))
# renumber the labels so that they conform to the specs
nlabels = np.amax(segmentation) + 1
renumber = np.zeros(nlabels, 'i')
for i, v in enumerate(lsort):
renumber[lines[v].label] = 0x010000 + (i + 1)
segmentation = renumber[segmentation]
# finally, output everything
print("writing lines")
if not os.path.exists(outputdir):
os.mkdir(outputdir)
lines = [lines[i] for i in lsort]
ocrolib.write_page_segmentation("%s.pseg.png" % outputdir, segmentation)
cleaned = ocrolib.remove_noise(binary, noise)
for i, l in enumerate(lines):
binline = psegutils.extract_masked(1 - cleaned,
l,
pad=pad,
expand=expand)
ocrolib.write_image_binary("%s/01%04x.bin.png" % (outputdir, i + 1),
binline)
# if args.gray:
# grayline = psegutils.extract_masked(
# gray, l, pad=args.pad, expand=args.expand)
# ocrolib.write_image_gray("%s/01%04x.nrm.png" % (outputdir, i+1), grayline)
print("%6d %s %4.1f %d" % (i, fname, scale, len(lines)))
# to proceed, we need a pseg file and a subdirectory containing text lines
assert os.path.exists(psegFile), "%s: no such file" % psegFile
assert os.path.isdir(inBase), "%s: no such directory" % inBase
# iterate through the text lines in reading order, based on the page segmentation file
pseg = ocrolib.read_page_segmentation(psegFile)
print("$$ %s=%s" % (psegFile, desc(pseg)))
regions = ocrolib.RegionExtractor()
print("$$ regions=%s" % regions)
regions.setPageLines(pseg)
im = Image.open(inFile)
print("~~%s %s" % (inFile, im.size))
print("$$ regions=%s=%s" % (regions, sorted(regions.__dict__)))
print("$$ regions.length=%s" % regions.length())
n = regions.length()
for i in range(1, n):
id = regions.id(i)
y0, x0, y1, x1 = regions.bbox(i)
# print("%5d: 0x%05X %s %d x %d" %
# (i, id, [y0, x0, y1, x1], y1 - y0, x1 - x0))
draw = ImageDraw.Draw(im)
draw.rectangle((x0, y0, x1, y1), outline=(255, 0, 0), width=3)
draw.rectangle((x0, y0, x1, y1), outline=(0, 0, 255), width=0)
# draw.rectangle((x0, y0, x1, y1), outline=255, width=5)
# draw.rectangle((x0, y0, x1, y1), outline=10, width=1)
del draw
# write output files
print("outFile=%s" % outFile)
im.save(outFile, "PNG")
print("outFile2=%s" % outFile2)
outDir2 = os.path.dirname(outFile2)
os.makedirs(outDir2, exist_ok=True)
im.save(outFile2, "PNG")
assert os.path.exists(outFile2)
# outFile3, _ = os.path.splitext(outFile)
# outFile3 = "%s.jpg" % outFile3
# print("outFile3=%s" % outFile3)
# im.save(outFile3, "JPEG")
# assert os.path.exists(outFile3)
return True
def textline(self, arg):
image = ocrolib.read_image_binary(arg)
height, width = image.shape
H = height
W = width
base, _ = ocrolib.allsplitext(arg)
base2 = os.path.splitext(arg)[0]
if not os.path.exists("%s/lines" % base):
os.system("mkdir -p %s/lines" % base)
#if os.path.exists(base2 + ".ts.png") :
# f = ocrolib.read_image_binary(base2 + ".ts.png")
# height, width = f.shape
# os.system("python "+args.libpath+"/anyBaseOCR-nlbin.py %s.pf.bin.png" % base2)
#else:
# os.system("python "+args.libpath+"/anyBaseOCR-nlbin.py %s" % arg)
#print("convert %s.ts.png %s/block-000.bin.png" % (base,base))
#os.system("convert %s.ts.png %s/block-000.bin.png" % (base,base))
#os.system("rm %s.bin.png %s.nrm.png" % (base, base))
file = open('%s/sorted_cuts.dat' % base, 'w')
l = "0 0 " + str(int(width)) + " " + str(
int(height)) + " 0 0 0 0\n"
file.write(l)
file.close()
#if not os.path.exists("%s/lines" % base) :
# os.system("mkdir %s/lines" % base)
blockarray = []
if os.path.exists(base + "/sorted_cuts.dat"):
blocks = open(base + "/sorted_cuts.dat", "r")
i = 0
for block in blocks:
words = block.split()
blockarray.append((int(words[0]), -int(words[1]),
int(words[2]), int(words[3]), i))
i += 1
else:
blockarray.append((0, 0, width, height, 0))
i = 0
j = 0
lines = []
for block in blockarray:
(x0, y0, x1, y1, i) = block
y0 = -y0
#blockImage = "%s/block-%03d" % (base, i)
os.system("convert %s.ts.png %s/temp.png" % (base, base))
img = Image.open("%s.ts.png" % base, 'r')
img_w, img_h = img.size
background = Image.new('RGBA', (W, H), (255, 255, 255, 255))
bg_w, bg_h = background.size
offX = (bg_w - img_w) // 2
offY = (bg_h - img_h) // 2
offset = (offX, offY)
background.paste(img, offset)
background.save("%s/temp.png" % base)
command = "python " + self.param[
'libpath'] + "/cli/anyBaseOCR-gpageseg.py %s/temp.png -n --minscale %f --maxlines %f --scale %f --hscale %f --vscale %f --threshold %f --noise %d --maxseps %d --sepwiden %d --maxcolseps %d --csminaspect %f --csminheight %f -p %d -e %d -Q %d" % (
base, self.param['minscale'], self.param['maxlines'], self.
param['scale'], self.param['hscale'], self.param['vscale'],
self.param['threshold'], self.param['noise'],
self.param['maxseps'], self.param['sepwiden'],
self.param['maxcolseps'], self.param['csminaspect'],
self.param['csminheight'], self.param['pad'],
self.param['expand'], self.param['parallel'])
if (self.param['blackseps']):
command = command + " -b"
if (self.param['usegauss']):
command = command + " --usegauss"
os.system(command)
pseg = ocrolib.read_page_segmentation("%s/temp.pseg.png" % base)
regions = ocrolib.RegionExtractor()
regions.setPageLines(pseg)
file = open('%s/sorted_lines.dat' % base, 'w')
for h in range(1, regions.length()):
id = regions.id(h)
y0, x0, y1, x1 = regions.bbox(h)
l = str(int(x0 - offX)) + " " + str(
int(img_h -
(y1 - offY))) + " " + str(int(x1 - offX)) + " " + str(
int(img_h - (y0 - offY))) + " 0 0 0 0\n"
file.write(l)
filelist = glob.glob("%s/temp/*" % base)
for infile in sorted(filelist):
os.system("convert %s %s/lines/01%02x%02x.bin.png" %
(infile, base, i + 1, j + 1))
lines.append("%s/lines/01%02x%02x.bin.png" %
(base, i + 1, j + 1))
j += 1
os.system("rm -r %s/temp/" % base)
os.system("rm %s/temp.png %s/temp.pseg.png" % (base, base))
i += 1
return lines