import numpy as np
import cv2
from matplotlib import pyplot as plt
import subprocess
import utils
import findCard
DEBUG = False
PHONEALIASES = ['ph.', 'phone', 'tel.']
def parseText(text):
text = text.strip()
# Replace long lines with regular lines
text = text.replace('\xe2\x80\x94', '-')
# Remove left single quotation mark
text = text.replace('\xe2\x80\x98','')
text = text.replace('\xef','i')
text = text.replace('\xc3\xa9', '6')
text = text.decode('utf-8', 'ignore')
return text
def checkBlur(img):
img = cv2.resize(img, (1000,600))
grayImg = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
grayImg = cv2.equalizeHist(np.copy(grayImg))
fft = np.fft.fft(grayImg)
avgFFT = np.average(fft)
threshFFT_x, threshFFT_y = np.where(fft> 1.25*avgFFT)
return len(threshFFT_x) > 130000
def getText(img, regions, debug=False):
textChunks = []
imgChunks = []
for region in regions:
x1, y1, x2, y2 = region
# Clip region and extract text
chunk = img[y1:y2, x1:x2, ::]
ret,chunk = cv2.threshold(chunk,140,255,0)
cv2.imwrite('tmp.jpg', chunk)
subprocess.call(['tesseract', 'tmp.jpg', 'tmp'])
f = open('tmp.txt')
lines = []
for line in f:
print line.strip()
if line.strip() != '':
lines.append(parseText(line))
textChunks.append(lines)
imgChunks.append(chunk)
f.close()
subprocess.call(['rm', 'tmp.jpg', 'tmp.txt'])
if DEBUG:
display = [(str(text), imgChunks[i]) for i, text in enumerate(textChunks)]
utils.display(display[:10])
utils.display(display[10:20])
return textChunks
def getRegions(img):
grayImg = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
# grayImg = cv2.equalizeHist(np.copy(grayImg))
edges = cv2.Canny(grayImg,100,200,apertureSize = 3)
if DEBUG:
utils.display([('Canny Edge Detection', edges)])
kernel = np.ones((3,3),np.uint8)
edges = cv2.dilate(edges,kernel,iterations = 14)
# edges = 255-edges
# utils.display([('', edges)])
contours, hierarchy = cv2.findContours(edges,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
if DEBUG:
utils.display([('Contours', edges)])
# Only take contours of a certain size
regions = []
for contour in contours:
imgH, imgW, _ = img.shape
[x, y, w, h] = cv2.boundingRect(contour)
if w < 50 or h < 50:
pass
elif w > .95*imgW or h > .95*imgH:
pass
else:
regions.append((x, y, x+w, y+h))
return regions
def drawRegions(img, regions):
for x1, y1, x2, y2 in regions:
cv2.rectangle(img, (x1,y1), (x2,y2), (0, 255, 0), 2)
if DEBUG:
utils.display([('Regions', img)])
return img
def processCard(img):
regions = getRegions(img)
text = getText(img, regions)
return regions, text
def guessFields(regions, textFields):
'''
Function to guess which fields are which based of of properties of the text
'''
def checkForPhone(line):
# Checks if any substrings of phone or related words are in the line
for word in PHONEALIASES:
if word in line:
return True
return False
def checkEmail(line):
if '@' in line and '.' in line:
return True
elif 'email' in line or 'e-mail' in line:
return True
else:
return False
def checkFax(line):
if 'fax' in line:
return True
else:
return False
def checkWebsite(line):
if 'www' in line:
return True
else:
return False
sizes = [(x2-x1)*(y2-y1) for x1,y1,x2,y2 in regions]
sortedSizes = sorted(sizes)
sortedSizes.reverse()
suggestedFields = []
for i in range(len(regions)):
suggestedLinesField = []
lines = textFields[i]
size = sizes[i]
lineCount = 0
for line in lines:
line = line.lower()
if size == sortedSizes[0] and lineCount == 0:
# Largest size suggest for company name
suggestedLinesField.append('Company')
elif checkForPhone(line):
suggestedLinesField.append('Phone')
elif checkFax(line):
suggestedLinesField.append('Fax')
elif checkWebsite(line):
suggestedLinesField.append('Website')
else:
suggestedLinesField.append('')
lineCount += 1
lineCount = 0
suggestedFields.append(suggestedLinesField)
return suggestedFields
if __name__ == "__main__":
imgs = utils.getImages('../../stanford_business_cards/photos/', 5)
# imgs = utils.getImages('../our_cards/', 8)
DEBUG = True
# img = utils.readImage('../../stanford_business_cards/photos/004.jpg')
# imgs = [('',img)]
# utils.display(imgs)
good, cardImg = findCard.findCard(imgs[4][1])
utils.display([('card',cardImg)])
regions, text = processCard(cardImg)
processedCard = drawRegions(cardImg, regions)
suggestedFields = guessFields(regions, text)
utils.display([('card',processedCard)])
# scores = []
# for imgName, img in imgs:
# # regions, text = processCard(img)
# # guessFields(regions, text)
# try:
# good, cardImg = findCard.findCard(img)
# except:
# good = False
# pass
# if good:
# scores.append((checkBlur(cardImg), imgName, cardImg))
# scores.sort()
# for score in scores:
# print score[0], score[1]
# imgs = [(str(score), img) for score, imgName, img in scores]
# for i in range(len(imgs)/10+1):
# utils.display(imgs[i*10:i*10+10])
