def checkOrientation(img, path):
print("Processing PDF Page: ", path)
errorIncurred = False
try:
pageDescrp = image_to_osd(img)
except TesseractError:
print("Tesseract Orientation Error")
errorIncurred = True
if not errorIncurred:
index1 = pageDescrp.find('Rotate')
index2 = pageDescrp.find('confidence')
orienDescrp = pageDescrp[index1:index2]
rotateAngle = list(filter(str.isdigit, orienDescrp))
orienDescrp = ""
for i in range(0, len(rotateAngle)):
orienDescrp = orienDescrp + rotateAngle[i]
# Perform Rotation if required
rotateAngle = int(orienDescrp)
if rotateAngle > 0:
rotatedImg = imutils.rotate_bound(img, rotateAngle)
cv2.imwrite(path, rotatedImg)
else: # Check Angular Rotation of Image:
checkSkewness(img, path)
else: # Check Angular Rotation of Image:
checkSkewness(img, path)
def rotate_img(image, count=0):
"""
rotate img if not aligned in right direction
"""
try:
text = pytesseract.image_to_osd(image)
except:
text = None
#print(text)
if text is not None and count < 4:
text = text.split('\n')
text = text[2].split(':')
rotate = int(text[1].strip())
if rotate == 90:
image = cv2.rotate(image, rotateCode=cv2.ROTATE_90_CLOCKWISE)
elif rotate == 270:
image = cv2.rotate(image,
rotateCode=cv2.ROTATE_90_COUNTERCLOCKWISE)
elif rotate == 180:
image = cv2.rotate(image, rotateCode=cv2.ROTATE_180)
elif text is None and count < 4:
count = count + 1
image = cv2.rotate(image, rotateCode=cv2.ROTATE_90_CLOCKWISE)
image = rotate_img(image, count)
return image
def read_card(encoded, orientation=0, algorithm='gbc', parser='regex'):
err = False
msg = None
image = cv2.imdecode(encoded, cv2.IMREAD_UNCHANGED)
print('before=>', image.shape)
obj_card = detect_object(image)
obj_card = deskew_object(obj_card)
prep = image_processing(obj_card)
if size_thresh(prep):
err = True
msg = {'error': f'gambar {image.shape} terlalu kecil'}
if not err and orientation:
osd = ts.image_to_osd(prep)
angle = re.search(r'(?<=Rotate: )\d+', osd).group(0)
if not image_orientation(angle):
err = True
msg = {'error': f'posisi gambar {angle} derajat'}
if not err:
data = ts.image_to_string(prep)
result = card_classifier(data, algorithm, parser)
return json.dumps(result)
return json.dumps(msg)
def detect(image):
# get bounding boxes
d = pytesseract.image_to_data(Image.open(image), output_type=Output.DICT)
# draw over original image the bounding boxes
img = cv2.imread(image)
n_boxes = len(d['level'])
for i in range(n_boxes):
(x, y, w, h) = (d['left'][i], d['top'][i], d['width'][i],
d['height'][i])
cv2.rectangle(img, (x, y), (x + w, y + h), (0, 255, 0), 2)
# convert cv2 to pillow image
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) # convert colorspace
bounding_boxes_image = Image.fromarray(img)
# parse ocr string
output_string = pytesseract.image_to_string(Image.open(image))
# get image osd metadata
osd = pytesseract.image_to_osd(Image.open(image), output_type=Output.DICT)
# print(f'osd: {osd}')
return {
'osd': osd,
'ocr_string': output_string,
}, bounding_boxes_image
def get_osd_info(img_path):
import pytesseract
'''
parse tesseract osd output string to a dict
for easier variable reading.
Info:
int Page_number
int Orientation_in_degrees
int Rotate
float Orientation_confidence
string Script
float Script_confidence
'''
osd_info = pytesseract.image_to_osd(img_path)
info = {}
for i in osd_info.split('\n'):
try:
if '.' in i.split(':')[1].strip():
info[str(i.split(':')[0].replace(' ', '_'))] = float(
i.split(':')[1].strip())
else:
info[str(i.split(':')[0].replace(' ', '_'))] = int(
i.split(':')[1].strip())
except:
info[str(i.split(':')[0].replace(' ',
'_'))] = i.split(':')[1].strip()
return info
def _detect_locale(item_rows: numpy.ndarray, locale: str) -> str:
"""Detects the right locale for the given items if required."""
if locale != 'auto':
# If locale is already specified, return as is.
return locale
# Convert to Pillow image and truncate overly long images.
image = Image.fromarray(item_rows[:9800, :])
try:
osd_data = pytesseract.image_to_osd(
image, output_type=pytesseract.Output.DICT)
except pytesseract.TesseractError:
return 'en-us'
possible_locales = SCRIPT_MAP.get(osd_data['script'])
assert possible_locales, 'Failed to automatically detect language.'
# If we can uniquely guess the language from the script, use that.
if len(possible_locales) == 1:
logging.info('Detected locale: %s', possible_locales[0])
return possible_locales[0]
# Otherwise, run OCR on the first few items and try to find the best matching locale.
item_names = run_ocr(item_rows[:30 * 35, :], lang='script/Latin')
def match_score_func(locale):
"""Computes how many items match for a given locale."""
item_db = _get_item_db(locale)
return sum(name in item_db for name in item_names)
best_locale = max(possible_locales, key=match_score_func)
logging.info('Detected locale: %s', best_locale)
return best_locale
def __extract_osd(self):
'''
Extracts a dict containing OSD (screen and dimensions) data for the image.
'''
self.__osd = pytesseract.image_to_osd(self.image,
lang=self.language,
output_type="dict")
def _get_angle_tesseract(
self,
config="--psm 12 --oem 3",
):
image_osd = pytesseract.image_to_osd(self._image, config=config)
rotation_angle = re.search('(?<=Rotate: )\d+', image_osd).group(0)
return rotation_angle
def img(url):
#if url.endswith(".jpeg") or url.endswith(".jpg"):
urlretrieve(url, "input.jpg")
im = Image.open(r"input.jpg")
img = prep(im)
cv2.imwrite("input.png", img)
image = cv2.imread("input.png")
#getting image in the correct orientation
try:
angle = int(
re.search('(?<=Orientation in degrees: )\d+',
pytesseract.image_to_osd(image)).group(0))
filename = "{}.png".format(os.getpid())
cv2.imwrite(filename, ndimage.rotate(image, angle))
#plt.imshow(ndimage.rotate(image, angle))
text = pytesseract.image_to_string(Image.open(filename),
config="-l eng --oem 1 --psm 3")
text = f1(text)
os.remove(filename)
return text
except TesseractError as e:
c = str(e.message)
text = pytesseract.image_to_string(Image.open("input.png"),
config="-l eng --oem 1 --psm 3")
text = f1(text)
#os.remove(filename)
print(c)
return text
def tesseract():
# If you don't have tesseract executable in your PATH, include the following:
#pytesseract.pytesseract.tesseract_cmd = r'<full_path_to_your_tesseract_exec$
# Example tesseract_cmd = r'C:\Program Files (x86)\Tesseract-OCR\tesseract'
# Simple image to string
#.encode('utf8') #add to print statement to make encoding work
print(pytesseract.image_to_string(Image.open(imgFile)).encode('utf8'))
# Get bounding box estimates
print(pytesseract.image_to_boxes(Image.open(imgFile)))
# Get verbose data including boxes, confidences, line and page numbers
print(pytesseract.image_to_data(Image.open(imgFile)))
# Get information about orientation and script detection
print(pytesseract.image_to_osd(Image.open(imgFile)))
# In order to bypass the internal image conversions, just use relative or ab$
# NOTE: If you don't use supported images, tesseract will return error
print(pytesseract.image_to_string(imgFile))
# get a searchable PDF
pdf = pytesseract.image_to_pdf_or_hocr(imgFile, extension='pdf')
# get HOCR output
hocr = pytesseract.image_to_pdf_or_hocr(imgFile, extension='hocr')
def image_to_osd(self, img):
"""画像から文字方向を読み込む"""
ocr_osd = pytesseract.image_to_osd(
img,
output_type=pytesseract.Output.DICT,
)
return ocr_osd
def img_oriented(image):
import re
osd = pytesseract.image_to_osd(image)
angle = osd.splitlines()[2].split()[1]
script = osd.splitlines()[4].split()[1]
print("angle: ", angle)
print("script: ", script)
(h, w) = image.shape[:2]
(cX, cY) = (w // 2, h // 2)
# grab the rotation matrix (applying the negative of the
# angle to rotate clockwise), then grab the sine and cosine
# (i.e., the rotation components of the matrix)
M = cv2.getRotationMatrix2D((cX, cY), -int(angle), 1.0)
cos = np.abs(M[0, 0])
sin = np.abs(M[0, 1])
# compute the new bounding dimensions of the image
nW = int((h * sin) + (w * cos))
nH = int((h * cos) + (w * sin))
# adjust the rotation matrix to take into account translation
M[0, 2] += (nW / 2) - cX
M[1, 2] += (nH / 2) - cY
# perform the actual rotation and return the image
return cv2.warpAffine(image, M, (nW, nH))
def de_skew(img, center = None, scale = 1.0,cropping= True):
tesData = pytesseract.image_to_osd(img) #tess angle for orientaion correction
pytesAngle = int(re.search('(?<=Rotate: )\d+', tesData).group(0))
tess_angle=360- pytesAngle
if tess_angle == 360:
tess_angle = 0
angle = getAngle(img) + tess_angle
print('Angle estimation arg for de_skew(): ',angle)
(h, w) = img.shape[:2]
center = (w // 2, h // 2)
rotation_mat = cv2.getRotationMatrix2D(center, angle, 1.0)
M = rotation_mat
abs_cos = abs(rotation_mat[0,0])
abs_sin = abs(rotation_mat[0,1])
# find the new width and height bounds
bound_w = int(h * abs_sin + w * abs_cos)
bound_h = int(h * abs_cos + w * abs_sin)
# subtract old image center (bringing image back to origo) and adding the new image center coordinates
rotation_mat[0, 2] += bound_w/2 - center[0]
rotation_mat[1, 2] += bound_h/2 - center[1]
if cropping:
openCVrotated = cv2.warpAffine(img.copy(), rotation_mat, (bound_w, bound_h),flags=cv2.INTER_CUBIC, borderMode=cv2.BORDER_WRAP)
else:
openCVrotated = cv2.warpAffine(img.copy(), M, center ,flags=cv2.INTER_CUBIC, borderMode=cv2.BORDER_WRAP)
cv2.imwrite(resFolderPath + r'\rotated.jpg',openCVrotated)
return openCVrotated
def orient_image(img):
try:
rotate = image_to_osd(img, output_type=Output.DICT)["rotate"]
# This tells it to use the
# highest quality interpolation algorithm that it has available,
# and to expand the image to encompass the full rotated size
# instead of cropping.
# The documentation does not say what color
# the background will be filled with.
# https://stackoverflow.com/a/17822099
angle = -float(rotate)
# if angle > 0:
# angle = 360 - angle
logger.info(f'Orientation angle {angle}')
k = angle // 90
if k != 0:
img = np.rot90(img, k=k)
# img = img.rotate(-rotate, resample=Image.BICUBIC, expand=True)
# sometimes an error can occur with tesseract reading the image
# maybe there is not enough text or dpi is not set
# this need to be handled
except Exception as e:
raise e
return img
def get_rotation_degree_by_quarter(self, image):
# Use tesseract to calculate the degree.
osd_info = pytesseract.image_to_osd(image, output_type=pytesseract.Output.DICT)
major_degree = osd_info['orientation']
print("Major Degree: {}".format(major_degree))
return major_degree
def skewImage3(image):
newdata = pytesseract.image_to_osd(image)
angle = re.search('(?<=Rotate: )\d+', newdata).group(0)
angle = int(angle)
if angle == 0:
return image, angle
return rotationImage(image, angle), angle
def get_best_text(image, iter: int):
rot_angle = 90 // iter
best_angle = None
best_conf = 0
for i in range(iter):
rot = image.rotate( i * rot_angle)
print("angle = {0}\n".format(i*rot_angle))
# rot = rot.convert("RGB")
# confs = pytesseract.image_to_data(rot, output_type=pytesseract.Output.DICT)["conf"]
rot.show()
# #convert non-integer entries to 0
# for i in range(len(confs)):
# # print(confs[i], " <- item\n")
# # print(type(confs[i]), " <- type\n")
# if type(confs[i]) != int:
# confs[i] = 0
output = pytesseract.image_to_osd(rot, output_type='dict')
confidence1 = output["script_conf"]
confidence2 = output["orientation_conf"]
# confidence = sum(confs)/len(confs)
# print("confidence = {0}, sum = {1}, len = {2}\n".format(confidence, sum(confs), len(confs)))
# print(pytesseract.image_to_osd(rot))
if confidence1 + confidence2 > best_conf:
best_angle = i*rot_angle
best_conf = confidence1 + confidence2
rot = image.rotate(best_angle)
return pytesseract.image_to_string(rot)
def fix_rotation(img):
rotated_img = img
# osd: orientation and script detection
tess_data = pytesseract.image_to_osd(img, nice=1)
angle = int(re.search(r"(?<=Rotate: )\d+", tess_data).group(0))
print("angle: " + str(angle))
if angle != 0 and angle != 360:
(h, w) = img.shape[:2]
center = (w / 2, h / 2)
# Perform the rotation
rotation_mat = cv2.getRotationMatrix2D(center, -angle, 1.0)
# Fixing the image cut-off by calculating the new center
abs_cos = abs(rotation_mat[0, 0])
abs_sin = abs(rotation_mat[0, 1])
bound_w = int(h * abs_sin + w * abs_cos)
bound_h = int(h * abs_cos + w * abs_sin)
rotation_mat[0, 2] += bound_w / 2 - center[0]
rotation_mat[1, 2] += bound_h / 2 - center[1]
rotated_img = cv2.warpAffine(img, rotation_mat, (bound_w, bound_h))
return rotated_img
def orientpage(filename):
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract'
f = open(filename, 'rb')
img_bytes = f.read()
f.close()
image = cv2.imdecode(np.frombuffer(img_bytes, dtype='uint8'),
cv2.IMREAD_COLOR) # Initially decode as color
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
gray = cv2.bitwise_not(gray)
rot_data = pytesseract.image_to_osd(image)
##print("[OSD] " + rot_data)
rot = re.search('(?<=Rotate: )\d+', rot_data).group(0)
angle = float(rot)
if angle > 0:
angle = 360 - angle
##print("[ANGLE] " + str(angle))
# rotate the image to deskew it
(h, w) = image.shape[:2]
center = (w // 2, h // 2)
M = cv2.getRotationMatrix2D(center, angle, 0.7)
rotated = cv2.warpAffine(image,
M, (w, h),
flags=cv2.INTER_CUBIC,
borderMode=cv2.BORDER_REPLICATE)
# TODO: Rotated image can be saved here
##print(pytesseract.image_to_osd(rotated));
cv2.imwrite(temp, rotated)
def pre_processing(img):
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'
#image_original = cv2.imread(img_file_path, cv2.IMREAD_COLOR)
#image_original = cv2.resize(image_original, (1700, 2200))
#image_original = cv2.resize(image_original, None ,fx=0.7, fy=0.7, interpolation= cv2.INTER_AREA)
image_scaled = cv2.resize(img,
None,
fx=2.0,
fy=2.0,
interpolation=cv2.INTER_LINEAR)
#### quay anh > 90
rotate_img = image_scaled
newdata = pytesseract.image_to_osd(image_scaled)
angle = 360 - int(re.search('(?<=Rotate: )\d+', newdata).group(0))
if angle > 0 and angle < 360:
rotate_img = rotate(image_scaled, angle)
# convert the image to grayscale
gray = cv2.cvtColor(rotate_img, cv2.COLOR_BGR2GRAY)
# threshold the image after Gaussian filtering
# medBlur = cv2.medianBlur(gray, 3)
# gauBlur = cv2.GaussianBlur(medBlur, (3,3), 10)
#return bit
thresh = cv2.threshold(gray, 90, 255,
cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
#kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,3))
#dilate = cv2.dilate(thresh, kernel, iterations=1)
return thresh
def pdf_language_detect(page_file):
try:
osd = pytesseract.image_to_osd(page_file)
language_script = osd.split('\nScript')[1][2:]
print('Language detected {0}'.format(language_script))
return language_script
except:
return None
def test_image_to_osd(test_file):
result = image_to_osd(test_file)
assert isinstance(result, unicode if IS_PYTHON_2 else str)
for key in [
'Page number', 'Orientation in degrees', 'Rotate',
'Orientation confidence', 'Script', 'Script confidence'
]:
assert key + ':' in result
def rotate(thres):
val = pytesseract.image_to_osd(thres)
if (val[50:53] == '270'):
thres = cv2.rotate(thres, cv2.ROTATE_90_COUNTERCLOCKWISE)
elif (val[51:53] == '90'):
thres = cv2.rotate(thres, cv2.ROTATE_90_CLOCKWISE)
elif (val[50:53] == '180'):
thres = cv2.rotate(thres, cv2.ROTATE_180)
return thres
def orientation_script_detection(self):
"""
Detects orientation and script of the file
:returns: angle and script
"""
image = cv2.imread(self.image)
osd = pytesseract.image_to_osd(image)
angle = re.search(r"(?<=Rotate: )\d+", osd).group(0)
script = re.search(r"(?<=Script: )\w+", osd).group(0)
return angle, script
def osd(image: Image) -> dict:
"""Returns orientation and script data for `image`.
"""
s = pytesseract.image_to_osd(image)
ret = dict()
for line in s.split('\n'): # type: ignore
if line:
key, value = line.split(':') # type: ignore
key, value = key.strip(), value.strip()
ret[key] = appropriate_type(value)
return ret
def deskewImageNew(img):
#Correct Skewness of image and return the image
cpy = img.copy()
try:
#Get text orientation from Tesseract
osd = pytesseract.image_to_osd(img)
rotationAngle = int(osd.split("\n")[2].split(":")[1].strip())
img1 = imutils.rotate_bound(img,rotationAngle)
return img1
except:
return cpy
def rotate(image):
im = Image.open(image)
try:
angle = 360 - int(
re.search('(?<=Rotate: )\d+',
pytesseract.image_to_osd(im)).group(0))
except:
angle = 0
im = im.rotate(angle)
im.save(path, "PNG")
def rotate_image(image, center = None, scale = 1.0):
angle=360-int(re.search('(?<=Rotate: )\d+', pytesseract.image_to_osd(image)).group(0))
(h, w) = image.shape[:2]
if center is None:
center = (w / 2, h / 2)
# Perform the rotation
M = cv2.getRotationMatrix2D(center, angle, scale)
rotated = cv2.warpAffine(image, M, (w, h))
return rotated
def check_orientation(image):
try:
orientation_details = pytesseract.image_to_osd(image)
angle = int(orientation_details.split("\n")[2].split(":")[-1])
if angle == 90:
image = cv2.rotate(image, cv2.cv2.ROTATE_90_CLOCKWISE)
elif angle == 180:
image = cv2.rotate(image, cv2.ROTATE_180)
return image
except:
return image
def deskewImageNew(img):
#Correct Skewness of image and return the image
try:
img1 = img.copy()
osd = pytesseract.image_to_osd(img1[:img1.shape[0] // 2,:])
rotationAngle = int(osd.split("\n")[2].split(":")[1].strip())
img1 = imutils.rotate_bound(img1,rotationAngle)
return img1
except Exception as e:
print("\tDeskewing Failed :-",e)
return img
from wand.image import Image as wand_Image
from PIL import Image as PIL_Image
from wand.color import Color
import os
import pytesseract
def build_images(force=False):
if not all([not force, os.path.isfile('./foo-0.png'), os.path.isfile('./foo-0.png')]):
all_pages = wand_Image(filename='./example.pdf', resolution=300)
for idx, page in enumerate(all_pages.sequence):
with Image(page) as i:
i.format = 'png'
i.background_color = Color('white')
i.alpha_channel = 'remove'
i.save(filename='foo-%s.png' % idx)
build_images()
boxes = pytesseract.image_to_boxes(PIL_Image.open('foo-0.png'), output_type=pytesseract.Output.DICT)
data = pytesseract.image_to_data(PIL_Image.open('foo-0.png'), output_type=pytesseract.Output.DICT)
osd = pytesseract.image_to_osd(PIL_Image.open('foo-0.png'), output_type=pytesseract.Output.DICT)
