Line detection with OpenCV Python and Hough transform

Question

I am trying to detect table lines and extract full table from an image with Python OpenCV and with Hough Transform algorithm. I need to have all coordinates of each line with the aim for draw the same table with same proportions. I understand theory how Hough transform works and tried to implement it without OpenCV, but it is very slow on big images.

Here is the code from example OpenCV Hough Transfrom

import cv2
import numpy as np

img = cv2.imread('image1.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray, 50, 150, apertureSize=3)
cv2.imshow("image", edges)
cv2.waitKey(0)
minLineLength = 100
maxLineGap = 10
lines = cv2.HoughLinesP(edges, 1, np.pi / 180, 50, minLineLength, maxLineGap)
for line in lines:
    for x1, y1, x2, y2 in line:
        cv2.line(img, (x1, y1), (x2, y2), (0, 255, 0), 2)

cv2.imwrite('houghlines5.jpg', img)

Canny edge detection returned an image Resulf of Canny edge detectione

But the result of detection is Resulf of Housh Transform

I do not know why Hough Transform left some lines of the table. Can you recommend something to do? Maybe another way to extract table from image? Thank you!

UPD. Original image original table needed to detect

A technique that is coming top off my head would be to count the number of black pixels in each row and column, the rows/columns with highest values are your target values, now you can use these results to get the intersection points, bounding boxes, etc. — ZdaR

kavko kavko · Accepted Answer · 2018-10-15T16:32:40

I think that you must prepare the image before you perform the HoughLinesP() for example get rid of the text, play with parameters or maybe dilate the edges etc. But if you would like to extract the template without text you can make your life easier by just drawing out all the text from this table (by searching small contours and making a white overlaying bounding box over the contours). In the example code I have made three steps: one by making template without Hough, second with HoughLines() and third with HoughLinesP(). Hope it helps a bit. Cheers!

Example:

import cv2
import numpy as np

### MAKING TEMPLATE WITHOUT HOUGH

# Read the image and make a copy then transform it to gray colorspace,
# threshold the image and search for contours.
img = cv2.imread('tablelines.png')
res = img.copy()
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
_, contours, hierarchy = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)

# Iterate through contours and draw a slightly bigger white rectangle
# over the contours that are not big enough (the text) on the copy of the image.
for i in contours:
    cnt = cv2.contourArea(i)
    if cnt < 500:
        x,y,w,h = cv2.boundingRect(i)
        cv2.rectangle(res,(x-1,y-1),(x+w+1,y+h+1),(255,255,255),-1)

# Display the result. Note that the image is allready the template!
cv2.imshow('res', res)
cv2.waitKey(0)
cv2.destroyAllWindows()

# Optional count the rows and columns of the table
count = res.copy()
gray = cv2.cvtColor(count, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY+cv2.THRESH_OTSU)
_, contours, hierarchy = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)

check = []
for i in contours:
    cnt = cv2.contourArea(i)
    if 10000 > cnt > 10:
        cv2.drawContours(count, [i], 0, (255,255,0), 2)
        M = cv2.moments(i)
        cx = int(M['m10']/M['m00'])
        cy = int(M['m01']/M['m00'])
        check.append([cx, cy])

check.sort(key = lambda xy: xy[1])
columns = 1

for i in range(0, len(check)-1):
    if check[i+1][1] + 5 >= check[i][1] >= check[i+1][1] - 5:
        columns += 1
    else:
        break
print(columns)

check.sort(key = lambda tup: tup[0])
rows = 1
for i in range(0, len(check)-1):
    if check[i+1][0] + 5 >= check[i][0] >= check[i+1][0] - 5:
        rows += 1
    else:
        break
print('Columns: ',columns)
print('Roiws : ',rows)

cv2.imshow('res', count)
cv2.waitKey(0)
cv2.destroyAllWindows()



### LINES WITH HOUGHLINES()

# Convert the resulting image from previous step (no text) to gray colorspace.
res2 = img.copy()
gray = cv2.cvtColor(res, cv2.COLOR_BGR2GRAY)

# You can either use threshold or Canny edge for HoughLines().
_, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
#edges = cv2.Canny(gray, 50, 150, apertureSize=3)

# Perform HoughLines tranform.  
lines = cv2.HoughLines(thresh,1,np.pi/180,200)
for line in lines:
    for rho,theta in line:
            a = np.cos(theta)
            b = np.sin(theta)
            x0 = a*rho
            y0 = b*rho
            x1 = int(x0 + 1000*(-b))
            y1 = int(y0 + 1000*(a))
            x2 = int(x0 - 1000*(-b))
            y2 = int(y0 - 1000*(a))

            cv2.line(res2,(x1,y1),(x2,y2),(0,0,255),2)

#Display the result.
cv2.imshow('res', res)
cv2.imshow('res2', res2)
cv2.waitKey(0)
cv2.destroyAllWindows()


### LINES WITH HOUGHLINESP()

# Convert the resulting image from first step (no text) to gray colorspace.
res3 = img.copy()
gray = cv2.cvtColor(res, cv2.COLOR_BGR2GRAY)

# Use Canny edge detection and dilate the edges for better result.
edges = cv2.Canny(gray, 50, 150, apertureSize=3)
kernel = np.ones((4,4),np.uint8)
dilation = cv2.dilate(edges,kernel,iterations = 1)

# Perform HoughLinesP tranform.  
minLineLength = 100
maxLineGap = 10
lines = cv2.HoughLinesP(dilation, 1, np.pi / 180, 50, minLineLength, maxLineGap)
for line in lines:
    for x1, y1, x2, y2 in line:
        cv2.line(res3, (x1, y1), (x2, y2), (0, 255, 0), 2)

#Display the result.
cv2.imwrite('h_res1.png', res)
cv2.imwrite('h_res2.png', res2)
cv2.imwrite('h_res3.png', res3)

cv2.imshow('res', res)
cv2.imshow('res2', res2)
cv2.imshow('res3', res3)
cv2.waitKey(0)
cv2.destroyAllWindows()

Result:

Columns: 7

Rows: 21

Line detection with OpenCV Python and Hough transform

2 Answers