OpenCV OMR Sheet - Detect Marked Answers in Python - Getting Proper Binarization

Question

While solving an OMR problem I am not completely able to detect all the marked answers correctly. Here is my input sheet.
Input Image

my code to Binarize the image.

import cv2

image = cv2.imread('input.png')
img = cv2.GaussianBlur(image,(5,5),0)
res, img = cv2.threshold(img, 60, 255, cv2.THRESH_BINARY)
img = 255 - img
cv2.imwrite('output.png',img)

Result I am getting.
Output generated

I have changed the gaussian blur and threshold parameters as follow.

 img = cv2.GaussianBlur(image,(7,7),0)  
 res, img = cv2.threshold(img, 90, 255, cv2.THRESH_BINARY)

The result I am getting after this change.
Result after changing Parameters

My Desired result should look something like this. Desired Result

Here is my complete code to detect the answers.

def solve(img,n_row = 50):
    height, width, channels = img.shape
    n_col = 4
    xShift = int(width/n_col)
    yShift = int(height/n_row)
    img = cv2.resize(img, (n_col * xShift, n_row*yShift))
    img = cv2.GaussianBlur(img,(5,5),0)
    res, img = cv2.threshold(img, 60, 255, cv2.THRESH_BINARY)
    img = 255 - img
    for row in range(0, n_row):
        tmp_img = img [row*yShift + 5:(row+1)*yShift - 5,]
        area_sum = []
        for col in range(n_col):
            area_sum.append(np.sum(tmp_img[1:,col*xShift :(col+1)*xShift]))                
        y = str(area_sum > np.median(area_sum) * 1)
        result.append(area_sum > np.median(area_sum) * 5)

Please If anyone can help me solving this problem I will be super thankful.

Answer Suggestions:
How can I Check out the white pixel count inside each bounding rectangle and keep only those bounding rectangles/contours whose area is above a minimum value.

inputImg= cv2.imread('input.jpg')
img = cv2.cvtColor(inputImg, cv2.COLOR_BGR2GRAY)    
mask = np.zeros(img.shape[:2], dtype=img.dtype)
ret, otsu_threshold = cv2.threshold(img, 120, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)     
contours, hierarchy = cv2.findContours(otsu_threshold, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
for c in contours:
    x,y,w,h = cv2.boundingRect(c)
    if cv2.contourArea(c) > 1500:
        cv2.rectangle(otsu_threshold, (x, y), (x+w, y+h), (0,255,0), 2)
cv2.imshow('Otsu', otsu_threshold)         
cv2.waitKey(0)      

Answer 1

Here is one way to do that in Python/OpenCV.

• Read the input
• Threshold on color
• Apply morphology to fill out the circles
• Get external contours
• For each contour, get the centroids and draw filled circles of radius=20 on the input at the centroid locations
• Save results

Input:

import cv2
import numpy as np

# read image
img = cv2.imread('omr_sheet.jpg')
h, w = img.shape[:2]

# trim 15 from bottom to remove partial answer
img = img[0:h-15, 0:w]

# threshold on color
lower=(120,60,80)
upper=(160,100,120)
thresh = cv2.inRange(img, lower, upper)

# apply morphology close
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (15,15))
morph = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel)
morph = cv2.morphologyEx(morph, cv2.MORPH_OPEN, kernel)

# get contours
result = img.copy() 
centers = []
contours = cv2.findContours(morph, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
contours = contours[0] if len(contours) == 2 else contours[1]
print("count:", len(contours))
print('')
i = 1
for cntr in contours:
    M = cv2.moments(cntr)
    cx = int(M["m10"] / M["m00"])
    cy = int(M["m01"] / M["m00"])
    centers.append((cx,cy))
    cv2.circle(result, (cx, cy), 20, (0, 255, 0), -1)
    pt = (cx,cy)
    print("circle #:",i, "center:",pt)
    i = i + 1
    
# print list of centers
#print(centers)

# save results
cv2.imwrite('omr_sheet_thresh.png',thresh)
cv2.imwrite('omr_sheet_morph.png',morph)
cv2.imwrite('omr_sheet_result.png',result)
# show results
cv2.imshow("thresh", thresh)
cv2.imshow("morph", morph)
cv2.imshow("result", result)

cv2.waitKey(0)
cv2.destroyAllWindows()

Threshold Image:

Morphology Image:

Results:

Answers:

count: 48

circle #: 1 center: (77, 3151)
circle #: 2 center: (78, 3087)
circle #: 3 center: (77, 3021)
circle #: 4 center: (76, 2959)
circle #: 5 center: (79, 2892)
circle #: 6 center: (77, 2830)
circle #: 7 center: (77, 2762)
circle #: 8 center: (78, 2695)
circle #: 9 center: (77, 2631)
circle #: 10 center: (76, 2566)
circle #: 11 center: (78, 2500)
circle #: 12 center: (77, 2435)
circle #: 13 center: (77, 2373)
circle #: 14 center: (77, 2301)
circle #: 15 center: (79, 2238)
circle #: 16 center: (79, 2178)
circle #: 17 center: (77, 2108)
circle #: 18 center: (78, 2045)
circle #: 19 center: (80, 1980)
circle #: 20 center: (78, 1913)
circle #: 21 center: (78, 1848)
circle #: 22 center: (80, 1786)
circle #: 23 center: (77, 1722)
circle #: 24 center: (77, 1657)
circle #: 25 center: (79, 1593)
circle #: 26 center: (79, 1524)
circle #: 27 center: (80, 1461)
circle #: 28 center: (77, 1395)
circle #: 29 center: (79, 1332)
circle #: 30 center: (76, 1265)
circle #: 31 center: (80, 1203)
circle #: 32 center: (73, 1136)
circle #: 33 center: (77, 1072)
circle #: 34 center: (80, 1007)
circle #: 35 center: (77, 944)
circle #: 36 center: (78, 878)
circle #: 37 center: (75, 815)
circle #: 38 center: (77, 747)
circle #: 39 center: (77, 684)
circle #: 40 center: (79, 618)
circle #: 41 center: (77, 554)
circle #: 42 center: (78, 488)
circle #: 43 center: (80, 423)
circle #: 44 center: (78, 359)
circle #: 45 center: (77, 293)
circle #: 46 center: (78, 232)
circle #: 47 center: (77, 165)
circle #: 48 center: (78, 102)

ADDITION:

This solution is probably less sensitive to color of ink. Just threshold on white using cv2.inRange() then invert.

import cv2
import numpy as np

# read image
img = cv2.imread('omr_sheet.jpg')
h, w = img.shape[:2]

# trim 15 from bottom and 5 from right to remove partial answer and extraneous red
img = img[0:h-15, 0:w-5]

# threshold on white color
lower=(225,225,225)
upper=(255,255,255)
thresh = cv2.inRange(img, lower, upper)
thresh = 255 - thresh

# apply morphology close
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (15,15))
morph = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel)
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (7,7))
morph = cv2.morphologyEx(morph, cv2.MORPH_CLOSE, kernel)

# get contours
result = img.copy() 
contours = cv2.findContours(morph, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
contours = contours[0] if len(contours) == 2 else contours[1]
print("count:", len(contours))
print('')
i = 1
for cntr in contours:
    M = cv2.moments(cntr)
    cx = int(M["m10"] / M["m00"])
    cy = int(M["m01"] / M["m00"])
    cv2.circle(result, (cx, cy), 20, (0, 255, 0), -1)
    pt = (cx,cy)
    print("circle #:",i, "center:",pt)
    i = i + 1
    
# save results
cv2.imwrite('omr_sheet_thresh2.png',thresh)
cv2.imwrite('omr_sheet_morph2.png',morph)
cv2.imwrite('omr_sheet_result2.png',result)

# show results
cv2.imshow("thresh", thresh)
cv2.imshow("morph", morph)
cv2.imshow("result", result)

cv2.waitKey(0)
cv2.destroyAllWindows()

Threshold Image:

Morphology Image:

Result:

Centers:

count: 48

circle #: 1 center: (78, 3150)
circle #: 2 center: (77, 3087)
circle #: 3 center: (78, 3020)
circle #: 4 center: (77, 2959)
circle #: 5 center: (78, 2892)
circle #: 6 center: (77, 2826)
circle #: 7 center: (77, 2760)
circle #: 8 center: (78, 2694)
circle #: 9 center: (79, 2630)
circle #: 10 center: (78, 2567)
circle #: 11 center: (78, 2501)
circle #: 12 center: (77, 2435)
circle #: 13 center: (75, 2373)
circle #: 14 center: (78, 2310)
circle #: 15 center: (78, 2240)
circle #: 16 center: (77, 2176)
circle #: 17 center: (78, 2106)
circle #: 18 center: (77, 2042)
circle #: 19 center: (79, 1979)
circle #: 20 center: (78, 1912)
circle #: 21 center: (78, 1848)
circle #: 22 center: (79, 1785)
circle #: 23 center: (78, 1722)
circle #: 24 center: (78, 1656)
circle #: 25 center: (78, 1589)
circle #: 26 center: (78, 1525)
circle #: 27 center: (79, 1458)
circle #: 28 center: (78, 1396)
circle #: 29 center: (78, 1328)
circle #: 30 center: (77, 1264)
circle #: 31 center: (78, 1201)
circle #: 32 center: (76, 1135)
circle #: 33 center: (78, 1071)
circle #: 34 center: (78, 1006)
circle #: 35 center: (77, 944)
circle #: 36 center: (77, 878)
circle #: 37 center: (77, 814)
circle #: 38 center: (77, 746)
circle #: 39 center: (78, 683)
circle #: 40 center: (77, 618)
circle #: 41 center: (77, 553)
circle #: 42 center: (78, 488)
circle #: 43 center: (78, 423)
circle #: 44 center: (77, 359)
circle #: 45 center: (76, 293)
circle #: 46 center: (79, 232)
circle #: 47 center: (75, 165)
circle #: 48 center: (79, 101)

Answer 2

Analysis:

The marked bubbles are in blue ink, which is a dominant color in the B-channel of the corresponding LAB color space.

• The B-channel captures blue and yellow colors in an image
• While A-channel captured red and green colors
• The L-channel highlights the brightness content in the image

To know more about LAB space go through this page

Code:

The following snippet shows how it is useful here:

img = cv2.imread('omr.jpg')

# convert to LAB space
lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
# store B-channel
b_channel = lab[:,:,2]

Notice how distinct the shaded bubbles are below:

# Otsu threshold on B-channel
th = cv2.threshold(b_channel,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)[1]

# morphological operations to close up the gaps within
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (5,5))
closing = cv2.morphologyEx(th, cv2.MORPH_CLOSE, kernel, iterations = 2)

# finding contours and drawing those above certain area onto black screen
contours, h = cv2.findContours(closing, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

black = np.zeros((img.shape[0], img.shape[1], 3), np.uint8)
for c in contours:
    area = cv2.contourArea(c)
    if area > 100:
        black = cv2.drawContours(black, [c], -1, (0, 255, 0), -1)

Answer 3

you can try this, it splits binary image into cells

and compares to find which cell has most while pixels in the row

https://dontrepeatyourself.org/post/bubble-sheet-multiple-choice-test-with-opencv-and-python/

but you must change the code follow your quiz form