def extract_text(self):
PDF_file = self.filename
out_folder_name = os.path.basename(self.filename)
if not os.path.exists(self.image_out_path):
os.mkdir(self.image_out_path)
if not os.path.exists(os.path.abspath(os.path.join(self.image_out_path,\
out_folder_name))):
os.mkdir(os.path.abspath(os.path.join(self.image_out_path,\
out_folder_name)))
index = 0
maxPages = pdf2image._page_count(PDF_file)
for page in range(0, maxPages, 10):
pages = pdf2image.convert_from_path(PDF_file,
dpi=200,
first_page=page,
last_page=min(
page + 10 - 1, maxPages))
for tpage in pages:
tpage.save(
os.path.abspath(
os.path.join(self.image_out_path, out_folder_name,
str(index) + ".jpg")), 'JPEG')
index = index + 1
print("Successfully saved images for each page for {}".format(
self.image_out_path))
english_text = list()
for filename in sorted(os.listdir(
os.path.join(self.image_out_path, out_folder_name)),
key=lambda x: int(os.path.splitext(x)[0])):
if filename.endswith("jpg"):
text = str(((pytesseract.image_to_string(
Image.open(
os.path.join(self.image_out_path, out_folder_name,
filename))))))
text = text.replace('-\n', '')
english_text.append(text)
corpus = " ".join(english_text)
corpus = re.sub(r'\n+', '\n', corpus).strip()
corpus = TextBlob(corpus)
for sentence in corpus.sentences:
self.english.append(sentence.string.replace("\n", " "))
print("English Text Extracted is : {}".format(self.english))
shutil.rmtree(self.image_out_path)
def read_pdf_as_image_to_list(path_file, dpi):
# print("images")
text_list = []
n_page = _page_count(path_file)
for i in range(1, n_page+1):
# print(i,text_list)
try:
pages = pdf2image.convert_from_path(path_file, dpi, first_page = i, last_page = i)
for page in pages:
text = str(pytesseract.image_to_string(page, lang="spa"))
text_list.append(text)
except Image.DecompressionBombError:
print('image size error')
text_list.append('')
pass
#print(text_list)
return text_list
def convertFlat(filename,
conditionLower,
conditionUpper,
batchSize,
progressBar,
outName="test",
pages=0,
pageOffset=0,
color=np.array([255, 255, 255]),
boundingBox=0):
'''
Iterate over selected pages of a pdf and change the color of all pixels withing a given range
'''
#save info of where to start
currentPageCounter = pageOffset
#get merger instance for outpdf creation
pdfMerger = PdfFileMerger()
#if pages weren't set we iterate over all
if pages == 0:
pages = _page_count(filename)
#variable for progress bar
totalPages = pages
X_Start, Y_Start, X_Len, Y_Len = 0, 0, 0, 0
#prepare coordinates for bounding box if it was created
if (len(boundingBox) == 2):
X_Start, Y_Start, X_Len, Y_Len = ConverCoordinates(boundingBox)
#work until nothing's left
while (pages > 0):
#use either batch size or w/e is left
if (pages > batchSize):
pagesToConvert = batchSize
else:
pagesToConvert = pages
#read pages
pageBuffer = convert_from_path(filename,
fmt='jpeg',
first_page=currentPageCounter + 1,
last_page=currentPageCounter +
pagesToConvert + 1)
for case in np.arange(pagesToConvert):
#get current page and convert to numpy array
im = pageBuffer[case]
fullPage = np.array(im)
#deal with potential bounding box
if (len(boundingBox) == 2):
editedPage = fullPage.copy()[Y_Start:Y_Start + Y_Len,
X_Start:X_Start + X_Len]
else:
editedPage = fullPage
#reshape for condition check
iterOrig = editedPage.reshape(
(editedPage.shape[0] * editedPage.shape[1], 3))
#prepare masks
maskUpper = np.all(conditionUpper <= iterOrig, axis=1)
maskLower = np.all(conditionLower >= iterOrig, axis=1)
#update all pixels that met condition
mask = np.logical_and(maskUpper, maskLower)
iterOrig[mask] = color
#restore shape
editedPage = iterOrig.reshape(
(editedPage.shape[0], editedPage.shape[1], 3))
#deal with potential bounding box
if (len(boundingBox) == 2):
fullPage[Y_Start:Y_Start + Y_Len,
X_Start:X_Start + X_Len] = editedPage
#resturn to previous format
im = Image.fromarray(fullPage)
#ugly code here due to time constraints/other priorities
im.save('temp\\temp.jpeg')
with open("temp\\tmp.pdf", "wb+") as f:
f.write(img2pdf.convert('temp\\temp.jpeg'))
with open("temp\\tmp.pdf", "rb") as f:
pdfMerger.append(f)
#update progress bar
progressBar.countChanged.emit(
int(((currentPageCounter + 1) / totalPages) * 100))
currentPageCounter += 1
#update loop termination var
pages -= pagesToConvert
#merge and save finished file
with open('result\\' + outName + ".pdf", 'wb') as fout:
pdfMerger.write(fout)
pdfMerger.close()
def convertAverage(filename,
filter,
batchSize,
progressBar,
outName="test",
pages=0,
pageOffset=0,
color=np.array([255, 255, 255]),
boundingBox=0):
#save info of where to start
currentPageCounter = pageOffset
#get merger instance for outpdf creation
pdfMerger = PdfFileMerger()
#if pages weren't set we iterate over all
if pages == 0:
pages = _page_count(filename)
#variable for progress bar
totalPages = pages
#read average calculated beforehand
averaged = np.array(Image.open(filter))
X_Start, Y_Start, X_Len, Y_Len = 0, 0, 0, 0
#prepare coordinates for bounding box if it was created
if (len(boundingBox) == 2):
X_Start, Y_Start, X_Len, Y_Len = ConverCoordinates(boundingBox)
averaged = averaged[Y_Start:Y_Start + Y_Len, X_Start:X_Start + X_Len]
#set correct shape
averaged = averaged.reshape((averaged.shape[0] * averaged.shape[1], 3))
#work until nothing's left
while (pages > 0):
#use either batch size or w/e is left
if (pages > batchSize):
pagesToConvert = batchSize
else:
pagesToConvert = pages
#read pages
pageBuffer = convert_from_path(filename,
fmt='jpeg',
first_page=currentPageCounter + 1,
last_page=currentPageCounter +
pagesToConvert + 1)
for case in np.arange(pagesToConvert):
#get current page and convert to numpy array
im = pageBuffer[case]
fullPage = np.array(im)
#deal with potential bounding box
if (len(boundingBox) == 2):
editedPage = fullPage.copy()[Y_Start:Y_Start + Y_Len,
X_Start:X_Start + X_Len]
else:
editedPage = fullPage
#reshape for condition check
iterOrig = editedPage.reshape((editedPage.shape[0] * editedPage.shape[1], 3))\
#calculate difference measure and aply to page
diff = np.sqrt(
np.power(np.sum(np.subtract(iterOrig, averaged), axis=1), 2))
mask = diff < 150
iterOrig[mask] = color
#restore shape
editedPage = iterOrig.reshape(
(editedPage.shape[0], editedPage.shape[1], 3))
#deal with potential bounding box
if (len(boundingBox) == 2):
fullPage[Y_Start:Y_Start + Y_Len,
X_Start:X_Start + X_Len] = editedPage
im = Image.fromarray(fullPage)
#ugly code here due to time constraints/other priorities
im.save('temp\\temp.jpeg')
with open("temp\\tmp.pdf", "wb+") as f:
f.write(img2pdf.convert('temp\\temp.jpeg'))
with open("temp\\tmp.pdf", "rb") as f:
pdfMerger.append(f)
#update progress bar
progressBar.countChanged.emit(
int(((currentPageCounter + 1) / totalPages) * 100))
currentPageCounter += 1
#update loop termination var
pages -= pagesToConvert
#merge and save finished file
with open('result\\' + outName + ".pdf", 'wb') as fout:
pdfMerger.write(fout)
pdfMerger.close()
def getAverageEstimate(filename,
batchSize,
progressBar,
outName="test",
pages=0,
pageOffset=0):
'''
Create estimated watermark by averaging over all pages in the document
'''
#save info of where to start
currentPageCounter = pageOffset
#if pages weren't set we iterate over all
if pages == 0:
pages = _page_count(filename)
totalPages = pages
#get zeroed out array for average calculation
firstPage = convert_from_path(filename,
fmt='jpeg',
first_page=1,
last_page=1)[0]
averaged = np.zeros_like(np.array(firstPage)).astype('uint64')
#address batch size
while (pages > 0):
if (pages > batchSize):
pagesToConvert = batchSize
else:
pagesToConvert = pages
#read a few pages to buffer
pageBuffer = convert_from_path(filename,
fmt='jpeg',
first_page=currentPageCounter + 1,
last_page=currentPageCounter +
pagesToConvert + 1)
for case in np.arange(pagesToConvert):
#get current page and convert to numpy array
im = pageBuffer[case]
fullPage = np.array(im)
#add current page
averaged = np.add(averaged, fullPage)
#update progress bar
progressBar.countChanged.emit(
int(((currentPageCounter + 1) / totalPages) * 100))
currentPageCounter += 1
#update loop termination var
pages -= pagesToConvert
#calculate average and save it
averaged = averaged / (currentPageCounter + 1)
im = Image.fromarray(averaged.astype('uint8'))
im.save("estimated_watermarks\\" + outName + '.jpeg')
return