cv2.connectedComponents doesn't work properly

Question

I want to use the function cv2.connectedComponents to connect components on a binary image, like the following...

.

Everything works, except the outputted labels array. In this array are only zeros and not sequential numbers as indicated, according to the identified components.

import cv2
import numpy as np

img = cv2.imread('eGaIy.jpg', 0)
img = cv2.threshold(img, 127, 255, cv2.THRESH_BINARY)[1]  # ensure binary
ret, labels = cv2.connectedComponents(img)

# Map component labels to hue val
label_hue = np.uint8(179*labels/np.max(labels))
blank_ch = 255*np.ones_like(label_hue)
labeled_img = cv2.merge([label_hue, blank_ch, blank_ch])

# cvt to BGR for display
labeled_img = cv2.cvtColor(labeled_img, cv2.COLOR_HSV2BGR)

# set bg label to black
labeled_img[label_hue==0] = 0

cv2.imshow('labeled.png', labeled_img)
cv2.waitKey()

outputted labels --> labels.shape: (256L, 250L)

[[0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 ..., 
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]]

I know what it is, I was just wandering if it was giving you 1 (only background label) or a normal number. I will test it to see if i can reproduce the error — api55, Jan 16 '18 at 13:36
It gives me the white components, not the 1 background component...Do you know, why the array only contains zeros? — freddykrueger, Jan 16 '18 at 13:38
I test it myself and it works, it gives me the same result as silencer, copy pasting your code.... if you print a numpy array you do not see all the values, where it has `...` means that you have a bunch of other values there. A simple test would be to do `np.sum(labels)` if it is 0 then it is only 0. Yo can try silencer suggestion for your test as well — api55, Jan 16 '18 at 13:44
A new "problem" occurs ... https://stackoverflow.com/questions/48303309/use-cv2-connectedcomponents-and-eliminate-elements-with-a-small-number-of-pixels — freddykrueger, Jan 17 '18 at 14:22

score 2 · Accepted Answer · answered Jan 16 '18 at 13:39

2

It works for me:

And you should be careful that the function only find the component of nonzero. In the source image, the components are the edges. And the returned are labeled image as the same size of source.

The output of

[[0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 ..., 
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]]

only represent the 4 corner regions(3x3) are all zeros, but it doesn't mean all elements are zeros.

If you call this after you call the cv2.connectedComponents:

print(set(labels.reshape(-1).tolist()))

You will get:

{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14}

It means there exist 14 components(edges), and 1 background(0).

answered Jan 16 '18 at 13:39

Kinght 金

17,681
4
60
74

I already know that the algorithm works visually. I have saved the array: "labels" as .csv using the algorithm "numpy. savetxt" and it contains only zeros. Why? – freddykrueger Jan 16 '18 at 13:47
It's ok for me when `np.savetxt("t.txt", labels)`, contains all 0~14 in float.Check again on your machine. Notice, almost all are zeros, not all are zeros. – Kinght 金 Jan 16 '18 at 14:03
"Almost all are zeros, not all are zeros." ... You are right! Thanks! – freddykrueger Jan 16 '18 at 14:06
Something like that, `0.000000000000000000e+00 1.000000000000000000e+00 0.000000000000000000e+00`. Not easy to find non-zero, but why not search `1~9` in the txt? – Kinght 金 Jan 16 '18 at 14:10
A new "problem" occurs ... https://stackoverflow.com/questions/48303309/use-cv2-connectedcomponents-and-eliminate-elements-with-a-small-number-of-pixels – freddykrueger Jan 17 '18 at 14:22
@freddykrueger Done! https://stackoverflow.com/questions/48303309/use-cv2-connectedcomponents-and-eliminate-elements-with-a-small-number-of-pixels – Kinght 金 Jan 17 '18 at 14:46

cv2.connectedComponents doesn't work properly

1 Answers1