CODEWITHSUNDEEP
pythonpython-3.xopencv
ivf1ze
ivf1ze

Reputation: 11

Minimal angled rectangle enclosing all black pixels of an image

I need to get the minimal angled rectangle enclosing all black pixels of an image.

Target images are scanned monochrome comic/manga pages that can be in any position (translated and rotated).

The desired result is similar to the following snippet when all black pixels are connected:

_, mono = cv2.threshold(image, 127, 255, cv2.THRESH_BINARY)
contours, _ = cv2.findContours(255-mono, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
rect = cv2.minAreaRect(contours[0])

In other words, I need to get the minAreaRect of the union of all contours, not only for one of them.

How can I do it using Python OpenCV?

Upvotes: 1

Views: 254

Answers (1)

HansHirse
HansHirse

Reputation: 18895

To keep your idea using the contours found by cv2.findContours, you just need to merge all of your contours, e.g. using np.vstack. Since you only need the plain coordinates for cv2.minAreaRect, merging them should be fine. To get the four vertices of the rotated rectangle, use cv2.boxPoints. Finally, drawing can be done with cv2.polylines.

Here's some short code snippet with a minimal example:

import cv2
import numpy as np

# Some test image
image = 255 * np.ones((300, 300), np.uint8)
cv2.circle(image, (100, 60), 30, 0, 1)
cv2.circle(image, (200, 200), 60, 0, 1)

# Find contours (with respect to OpenCV version); merge them
cnts = cv2.findContours(255-image, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
cnts = cnts[0] if len(cnts) == 2 else cnts[1]
cnts = np.vstack(cnts)

# Calculate minAreaRect of merged contours; determine points
pts = np.int32(cv2.boxPoints(cv2.minAreaRect(cnts)))
cv2.polylines(image, [pts], True, 192)

cv2.imshow('image', image)
cv2.waitKey(0)
cv2.destroyAllWindows()

That'd be the image:

Output

Hope that helps!

----------------------------------------
System information
----------------------------------------
Platform:    Windows-10-10.0.16299-SP0
Python:      3.8.1
NumPy:       1.18.1
OpenCV:      4.1.2
----------------------------------------

Upvotes: 4

Related Questions