def iter_images_and_pages(images):
"""This function iterates over a images and also the contained pages. As
OpenCV is not able to handle multipage TIFF files, we use the SDAPS internal
loading method for those."""
for filename in images:
try:
# Check whether this is a TIFF file (ie. try to retrieve the page count)
pages = image.get_tiff_page_count(filename)
is_tiff = True
except AssertionError:
pages = 1
is_tiff = False
for page in xrange(pages):
if not is_tiff:
img = cv2.imread(filename)
else:
# TIFF pages are zero based
surf = image.get_rgb24_from_tiff(filename, page, False)
width = surf.get_width()
height = surf.get_height()
stride = surf.get_stride()
# We need to ensure a sane stride!
np_width = stride / 4
# This converts by doing a copy; first create target numpy array
# We need a dummy alpha channel ...
target = np.empty((height, np_width), dtype=np.uint32)
tmp_surf = cairo.ImageSurface.create_for_data(
target.data, cairo.FORMAT_RGB24, width, height, stride)
cr = cairo.Context(tmp_surf)
cr.set_source_surface(surf)
cr.paint()
del cr
tmp_surf.flush()
del tmp_surf
# Now, we need a bit of reshaping
img = np.empty((height, width, 3), dtype=np.uint8)
# order should be BGR
img[:, :, 2] = 0xff & (target[:, :] >> 16)
img[:, :, 1] = 0xff & (target[:, :] >> 8)
img[:, :, 0] = 0xff & target[:, :]
yield img, filename, page
def iter_images_and_pages(images):
"""This function iterates over a images and also the contained pages. As
OpenCV is not able to handle multipage TIFF files, we use the SDAPS internal
loading method for those."""
for filename in images:
try:
# Check whether this is a TIFF file (ie. try to retrieve the page count)
pages = image.get_tiff_page_count(filename)
is_tiff = True
except AssertionError:
pages = 1
is_tiff = False
for page in xrange(pages):
if not is_tiff:
img = cv2.imread(filename)
else:
# TIFF pages are zero based
surf = image.get_rgb24_from_tiff(filename, page, False)
width = surf.get_width()
height = surf.get_height()
stride = surf.get_stride()
# We need to ensure a sane stride!
np_width = stride / 4
# This converts by doing a copy; first create target numpy array
# We need a dummy alpha channel ...
target = np.empty((height, np_width), dtype=np.uint32)
tmp_surf = cairo.ImageSurface.create_for_data(target.data, cairo.FORMAT_RGB24, width, height, stride)
cr = cairo.Context(tmp_surf)
cr.set_source_surface(surf)
cr.paint()
del cr
tmp_surf.flush()
del tmp_surf
# Now, we need a bit of reshaping
img = np.empty((height, width, 3), dtype=np.uint8)
# order should be BGR
img[:,:,2] = 0xff & (target[:,:] >> 16)
img[:,:,1] = 0xff & (target[:,:] >> 8)
img[:,:,0] = 0xff & target[:,:]
yield img, filename, page
def add_image(survey, file, duplex_scan=False, force=False, copy=True):
insert_dummy_pages, image_count_factor = _insert_dummy_pages(survey, duplex_scan)
if not check_image(survey, file, duplex_scan, force, message=True):
return
num_pages = image.get_tiff_page_count(file)
c = survey.questionnaire.page_count
if not insert_dummy_pages:
c = c * image_count_factor
if insert_dummy_pages:
c = c * image_count_factor
if copy:
tiff = survey.new_path('%i.tif')
shutil.copyfile(file, tiff)
else:
tiff = file
if copy:
tiff = os.path.basename(tiff)
else:
tiff = os.path.relpath(os.path.abspath(tiff), survey.survey_dir)
pages = list(range(num_pages))
while len(pages) > 0:
sheet = model.sheet.Sheet()
survey.add_sheet(sheet)
while len(pages) > 0 and len(sheet.images) < c:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = tiff
img.tiff_page = pages.pop(0)
# And a dummy page if required
if insert_dummy_pages:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = "DUMMY"
img.tiff_page = -1
img.ignored = True
def add_image(survey, file, duplex_scan=False, force=False, copy=True):
insert_dummy_pages, image_count_factor = _insert_dummy_pages(survey, duplex_scan)
if not check_image(survey, file, duplex_scan, force, message=True):
return
num_pages = image.get_tiff_page_count(file)
c = survey.questionnaire.page_count
if not insert_dummy_pages:
c = c * image_count_factor
if insert_dummy_pages:
c = c * image_count_factor
if copy:
tiff = survey.new_path("%i.tif")
shutil.copyfile(file, tiff)
else:
tiff = file
if copy:
tiff = os.path.basename(tiff)
else:
tiff = os.path.relpath(os.path.abspath(tiff), survey.survey_dir)
pages = range(num_pages)
while len(pages) > 0:
sheet = model.sheet.Sheet()
survey.add_sheet(sheet)
while len(pages) > 0 and len(sheet.images) < c:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = tiff
img.tiff_page = pages.pop(0)
# And a dummy page if required
if insert_dummy_pages:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = "DUMMY"
img.tiff_page = -1
img.ignored = True
def check_image(survey, file, duplex_scan=False, force=False, message=False):
insert_dummy_pages, image_count_factor = _insert_dummy_pages(survey, duplex_scan)
if not image.check_tiff_monochrome(file):
if message:
print(_('Invalid input file %s. You need to specify a (multipage) monochrome TIFF as input.') % (file,))
return False
num_pages = image.get_tiff_page_count(file)
c = survey.questionnaire.page_count
if not insert_dummy_pages:
c = c * image_count_factor
# This test is on the image count that needs to come from the file
if num_pages % c != 0 and not force:
if message:
print(_('Not adding %s because it has a wrong page count (needs to be a mulitple of %i).') % (file, c))
return False
return True
def check_image(survey, file, duplex_scan=False, force=False, message=False):
insert_dummy_pages, image_count_factor = _insert_dummy_pages(survey, duplex_scan)
if not image.check_tiff_monochrome(file):
if message:
print _("Invalid input file %s. You need to specify a (multipage) monochrome TIFF as input.") % (file,)
return False
num_pages = image.get_tiff_page_count(file)
c = survey.questionnaire.page_count
if not insert_dummy_pages:
c = c * image_count_factor
# This test is on the image count that needs to come from the file
if num_pages % c != 0 and not force:
if message:
print _("Not adding %s because it has a wrong page count (needs to be a mulitple of %i).") % (file, c)
return False
return True
def add_image(survey, file, duplex_scan=False, force=False, copy=True):
from sdaps import image
import shutil
# Insert dummy pages if the survey is duplex and the duplex option was not
# passed
if survey.defs.duplex:
# One image per questionnaire page in duplex mode
image_count_factor = 1
# No dummy pages in duplex mode
insert_dummy_pages = False
else:
# Two images per questionnaire page in duplex mode
image_count_factor = 2
# In simplex mode insertion of dummy pages depends on the command line
# optoin (default is True)
if duplex_scan:
insert_dummy_pages = False
else:
insert_dummy_pages = True
if not image.check_tiff_monochrome(file):
print _('Invalid input file %s. You need to specify a (multipage) monochrome TIFF as input.') % (file,)
raise AssertionError()
num_pages = image.get_tiff_page_count(file)
c = survey.questionnaire.page_count
if not insert_dummy_pages:
c = c * image_count_factor
# This test is on the image count that needs to come from the file
if num_pages % c != 0 and not force:
print _('Not adding %s because it has a wrong page count (needs to be a mulitple of %i).') % (file, c)
return
if insert_dummy_pages:
c = c * image_count_factor
if copy:
tiff = survey.new_path('%i.tif')
shutil.copyfile(file, tiff)
else:
tiff = file
if copy:
tiff = os.path.basename(tiff)
else:
tiff = os.path.relpath(os.path.abspath(tiff), survey.survey_dir)
pages = range(num_pages)
while len(pages) > 0:
sheet = model.sheet.Sheet()
survey.add_sheet(sheet)
while len(pages) > 0 and len(sheet.images) < c:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = tiff
img.tiff_page = pages.pop(0)
# And a dummy page if required
if insert_dummy_pages:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = "DUMMY"
img.tiff_page = -1
img.ignored = True
def iter_images_and_pages(images):
"""This function iterates over a images and also the contained pages. As
OpenCV is not able to handle multipage TIFF files, we use the SDAPS internal
loading method for those."""
for filename in images:
pages = 1
is_tiff = False
is_pdf = False
try:
# Check whether this is a TIFF file (ie. try to retrieve the page count)
pages = image.get_tiff_page_count(filename)
is_tiff = True
except AssertionError:
pass
if not is_tiff:
try:
gfile = Gio.File.new_for_path(filename)
pdf_doc = Poppler.Document.new_from_gfile(gfile, None, None)
pages = pdf_doc.get_n_pages()
is_pdf = True
except:
# Either not PDF/damaged or poppler not installed properly
pass
for page in xrange(pages):
if is_tiff:
# TIFF pages are zero based
surf = image.get_rgb24_from_tiff(filename, page, False)
img = to_opencv(surf)
elif is_pdf:
# Try to retrieve a single fullpage image, if that fails, render
# document at 300dpi.
THRESH = 10 #pt
pdfpage = pdf_doc.get_page(page)
page_width, page_height = pdfpage.get_size()
images = pdfpage.get_image_mapping()
if len(images) == 1 and (
abs(images[0].area.x1) < THRESH and
abs(images[0].area.y1) < THRESH and
abs(images[0].area.x2 - page_width) < THRESH and
abs(images[0].area.y2 - page_height) < THRESH):
# Assume one full page image, and simply use that.
surf = pdfpage.get_image(images[0].image_id)
else:
# Render page at 300dpi
surf = cairo.ImageSurface(cairo.FORMAT_RGB24, int(300 / 72 * page_width), int(300 / 72 * page_height))
cr = cairo.Context(surf)
cr.scale(300 / 72, 300 / 72)
cr.set_source_rgb(1, 1, 1)
cr.paint()
pdfpage.render_for_printing(cr)
del cr
img = to_opencv(surf)
else:
img = cv2.imread(filename)
yield img, filename, page
from sdaps import image
# Assume the first argument is a survey
survey = model.survey.Survey.load(sys.argv[1])
# We need the recognize buddies, as they are able to identify the data
from sdaps.recognize import buddies
# A sheet object to attach the images to
sheet = model.sheet.Sheet()
survey.add_sheet(sheet)
images = []
for file in sys.argv[2:]:
num_pages = image.get_tiff_page_count(file)
for page in xrange(num_pages):
images.append((file, page))
if len(images) == 0:
# No images, simply exit again.
sys.exit(1)
def add_image(survey, tiff, page):
img = model.sheet.Image()
survey.sheet.add_image(img)
# SDAPS assumes a relative path from the survey directory
img.filename = os.path.relpath(os.path.abspath(tiff), survey.survey_dir)
img.orig_name = tiff
img.tiff_page = page
def iter_images_and_pages(images):
"""This function iterates over a images and also the contained pages. As
OpenCV is not able to handle multipage TIFF files, we use the SDAPS internal
loading method for those."""
for filename in images:
pages = 1
is_tiff = False
is_pdf = False
try:
# Check whether this is a TIFF file (ie. try to retrieve the page count)
pages = image.get_tiff_page_count(filename)
is_tiff = True
except AssertionError:
pass
if not is_tiff:
try:
gfile = Gio.File.new_for_path(filename)
pdf_doc = Poppler.Document.new_from_gfile(gfile, None, None)
pages = pdf_doc.get_n_pages()
is_pdf = True
except:
# Either not PDF/damaged or poppler not installed properly
pass
for page in xrange(pages):
if is_tiff:
# TIFF pages are zero based
surf = image.get_rgb24_from_tiff(filename, page, False)
img = to_opencv(surf)
elif is_pdf:
# Try to retrieve a single fullpage image, if that fails, render
# document at 300dpi.
THRESH = 10 #pt
pdfpage = pdf_doc.get_page(page)
page_width, page_height = pdfpage.get_size()
images = pdfpage.get_image_mapping()
if len(images) == 1 and (
abs(images[0].area.x1) < THRESH and
abs(images[0].area.y1) < THRESH and
abs(images[0].area.x2 - page_width) < THRESH and
abs(images[0].area.y2 - page_height) < THRESH):
# Assume one full page image, and simply use that.
surf = pdfpage.get_image(images[0].image_id)
else:
# Render page at 300dpi
surf = cairo.ImageSurface(cairo.FORMAT_RGB24, int(300 / 72 * page_width), int(300 / 72 * page_height))
cr = cairo.Context(surf)
cr.scale(300 / 72, 300 / 72)
cr.set_source_rgb(1, 1, 1)
cr.paint()
pdfpage.render_for_printing(cr)
del cr
img = to_opencv(surf)
else:
img = cv2.imread(filename)
yield img, filename, page
def add_image(survey, file, duplex_scan=False, force=False, copy=True):
from sdaps import image
import shutil
# Insert dummy pages if the survey is duplex and the duplex option was not
# passed
if survey.defs.duplex:
# One image per questionnaire page in duplex mode
image_count_factor = 1
# No dummy pages in duplex mode
insert_dummy_pages = False
else:
# Two images per questionnaire page in duplex mode
image_count_factor = 2
# In simplex mode insertion of dummy pages depends on the command line
# optoin (default is True)
if duplex_scan:
insert_dummy_pages = False
else:
insert_dummy_pages = True
if not image.check_tiff_monochrome(file):
print _('Invalid input file %s. You need to specify a (multipage) monochrome TIFF as input.') % (file,)
raise AssertionError()
num_pages = image.get_tiff_page_count(file)
c = survey.questionnaire.page_count
if not insert_dummy_pages:
c = c * image_count_factor
# This test is on the image count that needs to come from the file
if num_pages % c != 0 and not force:
print _('Not adding %s because it has a wrong page count (needs to be a mulitple of %i).') % (file, c)
return
if insert_dummy_pages:
c = c * image_count_factor
if copy:
tiff = survey.new_path('%i.tif')
shutil.copyfile(file, tiff)
else:
tiff = file
if copy:
tiff = os.path.basename(tiff)
else:
tiff = os.path.relpath(os.path.abspath(tiff), survey.survey_dir)
pages = range(num_pages)
while len(pages) > 0:
sheet = model.sheet.Sheet()
survey.add_sheet(sheet)
while len(pages) > 0 and len(sheet.images) < c:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = tiff
img.tiff_page = pages.pop(0)
# And a dummy page if required
if insert_dummy_pages:
img = model.sheet.Image()
sheet.add_image(img)
img.filename = "DUMMY"
img.tiff_page = -1
img.ignored = True
def watch(cmdline):
# We need a survey that has the correct definitions (paper size, duplex mode)
# Assume the first argument is a survey
if os.path.exists('./WATCH/info'):
print('WATCH project found')
pass
else:
print('Creating WATCH project')
subprocess.call(['sdaps', 'setup', 'WATCH', './watch.tex'])
watchtexpath = (os.path.dirname(os.path.abspath(__file__)))
#loading dummy survey
print('Loading WATCH project')
survey = model.survey.Survey.load('WATCH')
# A sheet object to attach the images to
sheet = model.sheet.Sheet()
survey.add_sheet(sheet)
print('Listing all projects in ProjectsFolder')
#creating project dictionnary
surveyIdList = {}
#list of all subfolders containing 'info'
for file in Path(cmdline['projectsFolder']).walkfiles('info'):
s = file.dirname()
with open(s + '/info', "r") as infoFile:
#looking for survey id and add it to the dictionnary
lines = infoFile.read()
line = lines.split('\n')
for l in line:
words = l.split(' = ')
if words[0] == 'survey_id':
print('DETECT ! : ' + words[1])
surveyIdList[words[1]] = s
with open('surveyList.csv', 'w') as f:
for key in surveyIdList.keys():
f.write("%s,%s\n" % (key, surveyIdList[key]))
#file retrieval
print('Listing scanned files')
scans = os.listdir(cmdline['scanFolder'])
print(scans)
#temp folder creation
tempd = tempfile.mkdtemp()
print('Temp folder :' + tempd)
#folder with alreay processed scans
renamedFolder = cmdline['renamedFolder']
def is_tiff(scanned):
scan_title, scan_extension = os.path.splitext(scanned)
if scan_extension == '.tif' or scan_extension == '.tiff':
return True
else:
return False
def is_pdf(scanned):
scan_title, scan_extension = os.path.splitext(scanned)
if scan_extension == '.pdf':
return True
else:
return False
#convert and copy
for scan in scans:
scan_title, scan_extension = os.path.splitext(scan)
print(scan_title, scan_extension)
if is_pdf(scan):
print('PDF file found')
print('Scan title ' + scan_title,
'Scan extension ' + scan_extension)
tempscanpdf = tempfile.mktemp(suffix='.pdf', dir=tempd)
tempscantif = tempfile.mktemp(suffix='.tif', dir=tempd)
print('File', str(cmdline['scanFolder'] + '/' + scan),
'found, trying to convert to ' + tempscantif)
subprocess.call(
['cp', cmdline['scanFolder'] + '/' + scan, tempscanpdf])
print('Copied' + str(cmdline['scanFolder'] + '/' + scan) + 'to ' +
tempscanpdf)
#subprocess.call(['sdaps', 'add', "WATCH", tempscanpdf, '--convert'])
# for i, (img, filename, page) in enumerate(opencv.iter_images_and_pages(tempscanpdf)):
# print(img)
# print(filename)
# print(page)
scantoconvert = []
scantoconvert.append(tempscanpdf)
convert.convert_images(scantoconvert, tempscantif,
survey.defs.paper_width,
survey.defs.paper_height)
#subprocess.call(['pdfimages', '-tiff', cmdline['scanFolder']+'/'+scan, tempd+'/'+scan_title])
elif is_tiff(scan):
print('TIFF file found')
tempscantif = tempfile.mktemp(suffix='.tif', dir=tempd)
subprocess.call(
['cp', cmdline['scanFolder'] + '/' + scan, tempscantif])
else:
print('Wrong image format for file ' + scan)
#we retrieve all tiff to be processed
tiffscans = filter(is_tiff, os.listdir(tempd))
images = []
print('Files to be processed :' + str(tiffscans))
for file in tiffscans:
num_pages = image.get_tiff_page_count(tempd + '/' + file)
print(num_pages)
for page in range(num_pages):
images.append((tempd + "/" + file, page))
if len(images) == 0:
# No images, simply exit again.
sys.exit(1)
def add_image(survey, tiff, page):
img = model.sheet.Image()
survey.sheet.add_image(img)
# SDAPS assumes a relative path from the survey directory
img.filename = os.path.relpath(os.path.abspath(tiff),
survey.survey_dir)
img.orig_name = tiff
img.tiff_page = page
#print('Images added :'+str(img.filename)+str(img.orig_name)+str(img.tiff_page))
imgdummy = model.sheet.Image()
survey.sheet.add_image(imgdummy)
imgdummy.orig_name = "DUMMY"
imgdummy.filename = "DUMMY"
imgdummy.tiff_page = -1
imgdummy.ignored = True
#print('Images added :'+str(imgdummy.filename)+str(img.orig_name)+str(imgdummy.tiff_page))
while images:
# Simply drop the list of images again.
sheet.images = []
add_image(survey, *images.pop(0))
print('Adding image simplex mode')
if survey.defs.duplex:
print('Adding image duplex mode')
add_image(survey, *images.pop(0))
#print(images)
sheet.recognize.recognize()
#
for img in sheet.images:
if img.tiff_page != -1:
print(img.orig_name, img.tiff_page)
print('\tPage:', img.page_number)
print('\tRotated:', img.rotated)
print('\tMatrix (px to mm):', img.raw_matrix)
print('\tSurvey-ID:', sheet.survey_id)
print('\tGlobal-ID:', sheet.global_id)
print('\tBarcode-ID:', sheet.barcode_id)
print('\tQuestionnaire-ID:', sheet.questionnaire_id)
now = datetime.datetime.now()
datestamp = now.strftime('%Y%m%d%H%M%S%f')
tiffname = str(datestamp) + str(
sheet.questionnaire_id) + '_' + str(
sheet.survey_id) + '_' + str(sheet.barcode_id)
subprocess.call(['cp', img.orig_name, tiffname + ".tif"])
def iter_images_and_pages(images):
"""This function iterates over a images and also the contained pages. As
OpenCV is not able to handle multipage TIFF files, we use the SDAPS internal
loading method for those."""
for filename in images:
if not os.path.exists(filename):
raise IOError(errno.ENOENT, _("File does not exist"), filename)
pages = 1
is_tiff = False
is_pdf = False
try:
# Check whether this is a TIFF file (ie. try to retrieve the page count)
pages = image.get_tiff_page_count(filename)
is_tiff = True
except AssertionError:
pass
if not is_tiff:
try:
gfile = Gio.File.new_for_path(filename)
pdf_doc = Poppler.Document.new_from_gfile(gfile, None, None)
pages = pdf_doc.get_n_pages()
is_pdf = True
except:
# Either not PDF/damaged or poppler not installed properly
pass
for page in range(pages):
if is_tiff:
# TIFF pages are zero based
surf = image.get_rgb24_from_tiff(filename, page, False)
img = to_opencv(surf)
elif is_pdf:
# Try to retrieve a single fullpage image, if that fails, render
# document at 300dpi.
THRESH = 10 #pt
pdfpage = pdf_doc.get_page(page)
page_width, page_height = pdfpage.get_size()
images = pdfpage.get_image_mapping()
if len(images) == 1 and (
abs(images[0].area.x1) < THRESH
and abs(images[0].area.y1) < THRESH
and abs(images[0].area.x2 - page_width) < THRESH
and abs(images[0].area.y2 - page_height) < THRESH):
# Assume one full page image, and simply use that.
surf = pdfpage.get_image(images[0].image_id)
else:
dpi = 0
# Try to detect the DPI of the scan
for img in images:
if img.area.y2 - img.area.y1 < page_height / 2:
continue
surf = pdfpage.get_image(img.image_id)
# Calculate DPI from height
dpi_x = round(surf.get_height() /
(img.area.y2 - img.area.y1) * 72)
dpi_y = round(surf.get_width() /
(img.area.x2 - img.area.x1) * 72)
if abs(dpi_x - dpi_y) <= 1:
dpi = max(dpi, dpi_x, dpi_y)
# Fall back to 300dpi for odd values
if dpi < 199 or dpi > 601:
dpi = 300
surf = cairo.ImageSurface(cairo.FORMAT_RGB24,
int(dpi / 72 * page_width),
int(dpi / 72 * page_height))
cr = cairo.Context(surf)
cr.scale(dpi / 72, dpi / 72)
cr.set_source_rgb(1, 1, 1)
cr.paint()
pdfpage.render_for_printing(cr)
del cr
img = to_opencv(surf)
else:
img = cv2.imread(filename)
yield img, filename, page
def iter_images_and_pages(images):
"""This function iterates over a images and also the contained pages. As
OpenCV is not able to handle multipage TIFF files, we use the SDAPS internal
loading method for those."""
for filename in images:
if not os.path.exists(filename):
raise IOError(errno.ENOENT, _("File does not exist"), filename)
pages = 1
is_tiff = False
is_pdf = False
try:
# Check whether this is a TIFF file (ie. try to retrieve the page count)
pages = image.get_tiff_page_count(filename)
is_tiff = True
except AssertionError:
pass
if not is_tiff:
try:
gfile = Gio.File.new_for_path(filename)
pdf_doc = Poppler.Document.new_from_gfile(gfile, None, None)
pages = pdf_doc.get_n_pages()
is_pdf = True
except:
# Either not PDF/damaged or poppler not installed properly
pass
for page in range(pages):
if is_tiff:
# TIFF pages are zero based
surf = image.get_rgb24_from_tiff(filename, page, False)
img = to_opencv(surf)
elif is_pdf:
# Try to retrieve a single fullpage image, if that fails, render
# document at 300dpi.
THRESH = 10 #pt
pdfpage = pdf_doc.get_page(page)
page_width, page_height = pdfpage.get_size()
images = pdfpage.get_image_mapping()
if len(images) == 1 and (
abs(images[0].area.x1) < THRESH and
abs(images[0].area.y1) < THRESH and
abs(images[0].area.x2 - page_width) < THRESH and
abs(images[0].area.y2 - page_height) < THRESH):
# Assume one full page image, and simply use that.
surf = pdfpage.get_image(images[0].image_id)
else:
dpi = 0
# Try to detect the DPI of the scan
for img in images:
if img.area.y2 - img.area.y1 < page_height / 2:
continue
surf = pdfpage.get_image(img.image_id)
# Calculate DPI from height
dpi_x = round(surf.get_height() / (img.area.y2 - img.area.y1) * 72)
dpi_y = round(surf.get_width() / (img.area.x2 - img.area.x1) * 72)
if abs(dpi_x - dpi_y) <= 1:
dpi = max(dpi, dpi_x, dpi_y)
# Fall back to 300dpi for odd values
if dpi < 199 or dpi > 601:
dpi = 300
surf = cairo.ImageSurface(cairo.FORMAT_RGB24, int(dpi / 72 * page_width), int(dpi / 72 * page_height))
cr = cairo.Context(surf)
cr.scale(dpi / 72, dpi / 72)
cr.set_source_rgb(1, 1, 1)
cr.paint()
pdfpage.render_for_printing(cr)
del cr
img = to_opencv(surf)
else:
img = cv2.imread(filename)
yield img, filename, page
