OpenCV And Python: Connected Components Analysis

I have a working connected components analysis code working in C. It's actually a copy from the book 'Learning Opencv'. Now I am rewriting all that code to Python and I cannot find

Solution 1:

There is BSD license connected components code (in Cython) as part of scikit-image:

https://github.com/scikit-image/scikit-image/blob/master/skimage/measure/_ccomp.pyx

If you have the package installed, it is as simple as

from skimage import measure
import numpy as np

L = measure.label(image)
print "Number of components:", np.max(L)

Solution 2:

a bit dated of a reply but there's also this patch: http://code.opencv.org/attachments/467/opencv-connectedcomponents.patch

should be one of the faster implementations out there and still easy to call. should be integrated into mainline sometime in the future...

edit: it's been in mainline for sometime waiting on 3.0 to release. Don't ask me why they didn't release it earlier!

disclaimer - I'm the author :-)

Solution 3:

You should take a look at the documentation. As of OpenCV 2.2 there is a complete new interface for Python which covers all C/C++ functions :)

cv.FindContours should work for you :)

Solution 4:

OpenCV 3.0.0-dev has connectedComponents function but the doc says it is only for C++.

Solution 5:

In addition to Stefan van der Walt answer, here's a bit longer skimage labeling example:

from skimage import morphology
import cv2
import numpy as np
import sys


def get_image():
    size = (w, h) = (100, 100)
    img = np.zeros(size, np.uint8)
    cv2.rectangle(img, (10, 10), (19, 19), (128), -1)
    cv2.rectangle(img, (30, 20), (39, 39), (128), -1)
    cv2.rectangle(img, (40, 30), (49, 49), (128), -1)
    cv2.rectangle(img, (50, 70), (89, 79), (128), -1)
    return img


def show_image(img):
    cv2.imshow('result', img), cv2.waitKey(0)


if __name__ == '__main__':
    img = get_image()
    show_image(img)

    labels = morphology.label(img, background=0)
    label_number = 0
    while True:
        temp = np.uint8(labels==label_number) * 255
        if not cv2.countNonZero(temp):
            break
        show_image(temp)
        label_number += 1

    cv2.destroyAllWindows()