def default_freset(op, noise, boxes, img):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
#debug
cv2.imshow('ori', img)
cv2.waitKey()
pcs_obj = Preprocessing(img)
if op:
pcs_img = pcs_obj.natural_img_processing(noise)
else:
pcs_img = pcs_obj.digital_img_processing()
#debug
cv2.imshow('pre', pcs_img)
cv2.waitKey()
skew_obj = Deskew(pcs_img)
skew_img = skew_obj.run()
#debug
cv2.imshow('skew', skew_img)
cv2.waitKey()
tdt_obj = TextDetection(skew_img, img)
text_img = tdt_obj.detection()
#debug
flag = -1
for t_img, y, x in text_img:
#cv2.imshow('', t_img)
#cv2.waitKey()
t_img = cv2.copyMakeBorder(t_img, 10, 10, 10, 10, cv2.BORDER_CONSTANT, value=[255,255,255])
if flag <= y+1 and flag >= y-1:
tesseract_ocr(t_img, True)
else:
tesseract_ocr(t_img)
if boxes:
b_img = tesseract_boxes(t_img)
h, w = t_img.shape[:2]
gray[y*2:y*2+h, x*2:x*2+w] = b_img
flag = y
#debug
if boxes:
cv2.imshow('origin', tesseract_boxes(img))
cv2.imshow('result', gray)
cv2.waitKey()
def do_skew_correction(self,widget):
global imgloc
global skewcorrected
if imgloc == 0 :
errorwindow = builder.get_object("error_message_box")
errorwindow.set_transient_for(window)
errorwindow.set_markup("<b>No Image loaded to the application</b>")
errorwindow.format_secondary_markup("Add or Scan image to start skew correction")
errorwindow.show()
elif skewcorrected == 0 :
loc = imgloc
out=os.path.join(os.path.split(loc)[0],'skewcorrected'+os.path.splitext(os.path.split(loc)[1])[0]+os.path.splitext(os.path.split(loc)[1])[1])
d = Deskew(
input_file=loc,
display_image=None,
output_file=out,
r_angle=0)
d.run()
imgloc = out
img = builder.get_object("previmage")
img.set_from_file(imgloc)
skewcorrected = 1
successwindow = builder.get_object("error_message_box")
successwindow.set_transient_for(window)
successwindow.set_markup("<b>Skewness of the Image Corrected Successfully</b>")
successwindow.format_secondary_markup("You can now proceed Converting to text")
successwindow.show()
global zoomout
zoomout =1
else:
errorwindow = builder.get_object("error_message_box")
errorwindow.set_transient_for(window)
errorwindow.set_markup("<b>Image already Skew Corrected</b>")
errorwindow.format_secondary_markup("Further correction may increase the skewness")
errorwindow.show()
def process_image(img_path, output_path):
img = cv2.imread(img_path, cv2.IMREAD_GRAYSCALE)
print('Skew correcting image: ', img_path)
processed1 = process1(img)
cv2.imwrite(os.path.join(output_path, 'processed1.png'), processed1)
try:
angle = compute_skew(img)
print("deskew angle: {:.3f}".format(angle))
out = deskew_image(img, angle)
cv2.imwrite(os.path.join(output_path, 'processed2.png'), out)
except:
print('Failed')
d = Deskew(input_file=img_path,
display_image=False,
output_file=os.path.join(output_path, 'corrected.png'),
r_angle=0)
d.run()
def process_image(self):
#convert to grayscale and find canny edges
gray = cv2.imread(self.path)
edges = cv2.Canny(gray, 100, 150, apertureSize=3)
#get Hough Lines Probabilistic
minLineLength = 100
lines = cv2.HoughLinesP(image=edges,
rho=1,
theta=np.pi / 180,
threshold=120,
lines=np.array([]),
minLineLength=minLineLength,
maxLineGap=100)
if lines == []:
return
a, _, _ = lines.shape
x_diff = 0
y_diff = 0
for i in range(a):
x_diff += abs(lines[i][0][0] - lines[i][0][2])
y_diff += abs(lines[i][0][1] - lines[i][0][3])
x_diff = x_diff / len(lines)
y_diff = y_diff / len(lines)
print('x_diff: ', x_diff, 'y_diff: ', y_diff)
if x_diff > y_diff:
ratio = x_diff / y_diff
print('Image is Horizontal, ratio is:', x_diff / y_diff)
else:
ratio = y_diff / x_diff
print('Image is Vertical, ratio is:', y_diff / x_diff)
return
#self.rotate_image()
if ratio < 7.0:
print('we need to deskew this image')
deskew_obj = Deskew(self.path)
deskew_obj.deskew()
from time import sleep from cv2.cv2 import * import numpy as np from matplotlib import pyplot as plt from deskew import Deskew path = "C:\\Users\\OmarMarridi\\Desktop\\Semester\\Digtal Image Processing\\ImageProject\\Preprocessing DataSet\\" deskew_obj = Deskew(path+'Doc14.tif', 'Yes', 0) image=deskew_obj.run() image = threshold(image,130,255,THRESH_BINARY) imwrite(path+'testborder.tif',image[1])
from skew_detect import SkewDetect
from deskew import Deskew
from text_cleaner import clean_image
img_path = './images/12.png'
clean_image(img_path, './')
img_path = 'cleaned.png'
#sd = SkewDetect(input_file=img_path, output_file='./output.txt', display_output='Yes')
#sd.run()
d = Deskew(input_file=img_path,
display_image=False,
output_file='output_img.png',
r_angle=0)
d.run()
from detect_blur import *
from deskew import Deskew
path = "C:\\Users\\OmarMarridi\\Desktop\\Semester\\Digtal Image Processing\\ImageProject\\Preprocessing DataSet\\"
files = [
"Doc07.tif", "Doc08.tif", "Doc09.tif", "Doc011.tif", "Doc11.tif",
"Doc12.tif", "Doc13.tif", "Doc14.tif", "Doc15.tif", "Doc077.tif",
"Fax001.tif", "Fax003.tif", "wordlist12.tif", "wordlist23.tif",
"Document001.tif", "Doc066.tif"
]
for i in range(0, files.__len__() - 1, 1):
print("Processing Image \"" + files[i] + "\"")
orig = imread(path + files[i], IMREAD_GRAYSCALE)
img = orig
deskew_obj = Deskew(img, 'Yes', 0)
image = deskew_obj.run()
_, image = threshold(image, 130, 255, THRESH_BINARY)
_, thresh = threshold(image, 1, 255, THRESH_BINARY)
contours = findContours(image, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE)
cnt = contours[0]
x, y, w, h = boundingRect(cnt)
image = image[y:y + h, x:x + w]
image = equalizeHist(image)
#if(detectBlur(img, 100)):
#print("Blurred Image")
blur = medianBlur(image, 3)
for k in range(1, 5, 1):
blur = medianBlur(blur, 3)
flags=cv2.CASCADE_SCALE_IMAGE)
# if a face find in this case we consider this face and image and continue our job
if (len(faces) != 0):
M = cv2.getRotationMatrix2D(center, 180, 1.0)
docImage = cv2.warpAffine(docImage, M, (width, height))
print("Found {0} faces!".format(len(faces)))
return docImage, faces
#################################################################
#################################################################
inputAddress = 'patent3.jpg'
outputAddress = 'patent3Ro.jpg'
des = Deskew(inputAddress, False, outputAddress, 0)
des.run()
image = Image.open(outputAddress)
box = getbox(image)
result = image.crop(box)
result.save('sampleCut.jpg')
# Read the image
image = cv2.imread('sampleCut.jpg')
image, faces = detectFace(image)
# we keep only the biggest face. because sometime some noise fool the algorithm
faceArea = 0
faceX = 0
faceY = 0