detect filled rectangles in image - c++

Question

How to detect filled rectangles in image?
I need to get the bounding box for the 4 white (filled with white) rectangles in the right side of the image, but not the big rectangle in the middle with a white outline

Answer 1

You can isolate each contour by drawing the contour on a mask. Then you can use that mask on the image to calculate the average color. A high average indicates that the contour contains mostly white, so it is likely a contour you want.
Result:
Code:
import numpy as np
import cv2
#load the image
img = cv2.imread("form.png")
# create grayscale
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
#Find contours (external only):
im, contours, hierarchy = cv2.findContours(gray, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
#draw contours on original image
for cnt in contours:
# disregard small contours cause by logo and noise
if cv2.contourArea(cnt) > 10000:
#isolate contour and calculate average pixel value
mask = np.zeros(gray.shape[:2],np.uint8)
cv2.drawContours(mask,[cnt],0,255,-1)
mean_val = cv2.mean(gray,mask = mask)
# a high value indicates the contour contains mostly white, so draw the contour (I used the boundingRect)
if mean_val[0] > 200:
x,y,w,h = cv2.boundingRect(cnt)
cv2.rectangle(img, (x,y),(x+w,y+h), (0,0,255), thickness=4)
# show/save image
cv2.imshow("Image", mask)
cv2.imwrite("result.jpg", img)
cv2.waitKey(0)
cv2.destroyAllWindows()
Note: you can also load the image as grayscale and skip creating one, but I used it here so I could draw more obvious red boxes.
Also be aware the code given might not generalize well, but it shows the concept.