How To Obtain Boundary Coordinates Of Binary Mask With Holes?

I have the following image: I would like to obtain a list with (x, y)-coordinates of the outer and inner contour for each blob (let's call them blob A and B). import cv2 from skim

Solution 1:

You get the (x, y)-coordinates directly from cv2.findContours. To identify the single blobs, have a look at the hierarchy hier. The fourth index tells you, to which outer (or parent) contour a possible inner (or child) contour is related. Most outer contours have an index of -1, all other have non-negative values. So, for plotting/drawing, a naive approach would be, while iterating the contours, to increase a blob counter every time you see a -1, and draw all contours with the same color until the next -1 shows.

import cv2
from skimage import io         # Only needed for web grabbing images, use cv2.imread for local images# Read image; find contours with hierarchy
blob = io.imread('https://i.stack.imgur.com/Ga5Pe.png')
contours, hier = cv2.findContours(blob, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)

# Define sufficient enough colors for blobs
colors = [(255, 0, 0), (0, 255, 0), (0, 0, 255)]

# Draw all contours, and their children, with different colors
out = cv2.cvtColor(blob, cv2.COLOR_GRAY2BGR)
k = -1for i, cnt inenumerate(contours):
    if (hier[0, i, 3] == -1):
        k += 1
    cv2.drawContours(out, [cnt], -1, colors[k], 2)

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

Of course, obtaining all contours belonging to the same blob can be optimized using NumPy, but looping feels most intuitive here. I omitted all the other stuff (skimage, Matplotlib), since they didn't seem relevant here. As I said, the (x, y)-coordinates are already stored in contours.

Hope that helps!

---

EDIT: I haven't validated, if OpenCV always obtains all contours belonging to one most outer contour continously, or if - for example - all contours for a given hierarchy level are stored subsequently. So, for more complicated hierarchies, this should be tested beforehand, or the mentioned index finding using NumPy should be used right from the start.