def get_data_from_image(self, image):
try:
pytesseract.tesseract_cmd = self.sys_cfg['tesseract_directory']
self.bounding_boxes = pytesseract.image_to_boxes(image)
self.redraw(False)
return pytesseract.image_to_string(image)
except pytesseract.TesseractNotFoundError:
self.set_status('Did not find tesseract.exe at specified directory', 'red')
def processImage(filename):
img = cv2.cvtColor(cv2.imread(filename), cv2.COLOR_BGR2GRAY)
# alpha = 2.5
# beta = 70
# for y in range(img.shape[0]):
# for x in range(img.shape[1]):
# for c in range(img.shape[2]):
# img[y,x,c] = np.clip(alpha*img[y,x,c] + beta, 0, 255)
d = pt.image_to_boxes(img, output_type=Output.DICT)
return d
ap = argparse.ArgumentParser()
ap.add_argument("-i",
"--image",
required=True,
help="path to input image to be OCR'd")
ap.add_argument("-t",
"--training-dir",
required=False,
default=TRAINING_BOXES_DIR,
help="path to input image to be OCR'd")
ap.add_argument("-p",
"--process",
required=False,
default=0,
help="path to input image to be OCR'd")
args = vars(ap.parse_args())
image_path = args['image']
filename = image_path.split("/")[-1]
workdir = image_path.split("/")[:-1]
workdir = "/".join(workdir) + "/"
if args['process']:
image = binarize_image(workdir, filename)
else:
image = Image.open(image_path)
boxes = pytesseract.image_to_boxes(image, config=CONFIG_TESSERACT, lang=LANG)
with open(TRAINING_BOXES_DIR + filename.split(".")[0] + ".boxes",
"w+") as boxesfile:
boxesfile.write(boxes)
boxesfile.close()
hocr = open("output.hocr", "r", encoding="utf-8").read()
#extract coordinate information from hocr
soup = BeautifulSoup(hocr, "html.parser")
words = soup.find_all('span',class_='ocrx_word')
word_coordinates = get_coordinates(words)
lines = soup.find_all('span',class_='ocr_line')
line_coordinates = get_coordinates(lines)
paragraphs = soup.find_all('p',class_='ocr_par')
paragraph_coordinates = get_coordinates(paragraphs)
# Draw the bounding box
img = cv2.imread(filename)
boxes = pt.image_to_boxes(img)
h, w, _ = img.shape
for c in word_coordinates:
img = cv2.rectangle(img,(int(c[0]),int(c[1])),(int(c[2]),int(c[3])),(255,0,0),2) #blue
for c in line_coordinates:
img = cv2.rectangle(img,(int(c[0]),int(c[1])),(int(c[2]),int(c[3])),(0,0,255),2) #red
for c in paragraph_coordinates:
img = cv2.rectangle(img,(int(c[0]),int(c[1])),(int(c[2]),int(c[3])),(255,255,0),2) #cyan
for b in boxes.splitlines():
b = b.split(' ')
img = cv2.rectangle(img, (int(b[1]), h - int(b[2])), (int(b[3]), h - int(b[4])), (0, 255, 0), 2) #green
import cv2
from pytesseract import pytesseract
pytesseract.tesseract_cmd = "C:\\Program Files\\Tesseract-OCR\\tesseract.exe"
img = cv2.imread("ur_mom.png")
height, width, c = img.shape
letter_boxes = pytesseract.image_to_boxes(img)
print(letter_boxes)
for box in letter_boxes.splitlines():
box = box.split()
print(box)
x,y,w,h = int(box[1]),int(box[2]),int(box[3]),int(box[4])
cv2.rectangle(img,(x,height-y),(w,height-h), (0,0,255),3)
cv2.putText(img,box[0],(x,height-h+32),cv2.FONT_HERSHEY_COMPLEX,1,(0,255,0),2)
cv2.imwrite('eeeeee.png', img)
cv2.imshow("window", img)
cv2.waitKey(0)
def waterFill(image, charMap, color=False, blur=False):
h, w = image.shape
if blur:
image = cv2.GaussianBlur(image, ksize=(3, 3), sigmaX=2)
boxes = tesseract.image_to_boxes(image,
lang="eng",
config=r'--oem 3 --psm 3')
up_thres = 0.032
down_thres = 0.095
left_thres = 0
right_thres = 0.012
boxed_image = image.copy()
if color: boxed_image = cv2.cvtColor(boxed_image, cv2.COLOR_GRAY2BGR)
# plt.imshow(boxed_image)
# plt.waitforbuttonpress(0)
# plt.close()
upright = inverted = 0
for b in boxes.splitlines():
# Letter, coordinates of bottom left corner of bounding box,
# coordinates of top right of bounding box
bx = b.split(" ")
#If tesseract identifies it as symmetric letter, continue
if bx[0] in charMap.loc[:, "CharName"].values:
rw = charMap[charMap["CharName"] == bx[0]].first_valid_index()
if charMap.loc[rw, "Symmetric"] == True: #Symmetric column
continue
bx[1:] = map(int, bx[1:]) #convert coordinates to integers
#get the character found by tesseract
cropped_image = image[h - bx[4]:h - bx[2] + 1, bx[1]:bx[3] + 1]
#convert any blur to either black or white pixel values
t = 143
cropped_image_b = binarize_image(cropped_image.copy(), threshold=t)
cropped_image_b = np.abs(cropped_image_b - 255) #invert the image
#Calculate capacities for character
cib_h, cib_w = cropped_image_b.shape
cib_area = cib_h * cib_w
#Total capacity, not capacity/pixel
U, D, L, R, _, _ = water_fill6(cropped_image_b)
#Capacity per pixel
try:
U, D, L, R = U / cib_area, D / cib_area, L / cib_area, R / cib_area
except ZeroDivisionError:
U, D, L, R = 0, 0, 0, 0
if U < up_thres: U = 0
if D < down_thres: D = 0
if L < left_thres: L = 0
if R < right_thres: R = 0
upright += D + R
inverted += U + L
#output image with boxes around tesseract detected characters
boxed_image = cv2.rectangle(boxed_image, (bx[1], h - bx[2]),
(bx[3], h - bx[4]), (80, 176, 0), 0)
if upright >= 2.5 * inverted:
return "upright", boxed_image, upright, inverted
else:
return "inverted", boxed_image, upright, inverted