Detecting blobs on image with OpenCV

Question

I want to get some descriptors for each white area on image to filter that areas and operate with them separately. How can I do it?

I have read How to use OpenCV SimpleBlobDetector and http://www.learnopencv.com/blob-detection-using-opencv-python-c/ but still can't get any result with my simple image.

Here is my code in python

img = cv2.imread("map.jpg", cv2.IMREAD_GRAYSCALE)
params = cv2.SimpleBlobDetector_Params()
params.blobColor = 255
params.filterByColor = True
params.minArea = 16
params.filterByArea = True
detector = cv2.SimpleBlobDetector_create(params)
keypoints = detector.detect(255 - img)
len(keypoints)
# 0

OpenCV 3.1.0

Image is grayscaled.

UPD: Code updated following comment by @api55

you need to pass the blob params, setting the blobColor to 255 and filter by color to true. By default it tries to find dark blobs in white background. Or you could invert the colors of the image — api55, May 11 '16 at 12:33
@api55 I have updated code (in post too) but still no result — Askhat Nuriddinov, May 11 '16 at 12:43
Try changing also the minArea parameter to something small, like 16 (4x4 pixel blob approx). In the tutorial, it says that filter by color is not working (not sure what version you have).... you can try inverting the image colors — api55, May 11 '16 at 12:51
@Antonio The result is an empty array of keypoints. What should I show? — Askhat Nuriddinov, May 11 '16 at 13:06
Check [the other answer](http://stackoverflow.com/a/28573944/2436175), seems more up to date. — Antonio, May 11 '16 at 13:07
Possible duplicate of [Tutorial on OpenCV SimpleBlobDetector](http://stackoverflow.com/questions/8076889/tutorial-on-opencv-simpleblobdetector) — Antonio, May 11 '16 at 13:13
try loading the image as greyscale, and remove the inverted color thing... `cv2.imread("map.jpg", cv2.IMREAD_GRAYSCALE)` — api55, May 11 '16 at 13:18
@api55 I tried just to copy the code and change `imread("blob.jpg")` to `imread("map.jpg")` and add `print(len(keypoints))` and got the same result. — Askhat Nuriddinov, May 11 '16 at 13:21
@AskhatNuriddinov even if your image is greyscale, OpenCV may load it as BGR... don't ask me why, but I it seems that is the default.... — api55, May 11 '16 at 13:31
@Antonio The image is showed properly with `matplotlib.pyplot.subplot`. Converting it to bmp didn't solve my problem. — Askhat Nuriddinov, May 11 '16 at 13:34
@api55 Sorry, haven't updated the code in the post. On my computer it was grayscaled. — Askhat Nuriddinov, May 11 '16 at 13:36
Please try the (updated) code [here](https://github.com/spmallick/learnopencv/blob/master/BlobDetector/blob.py) — Antonio, May 11 '16 at 13:41

score 1 · Answer 1 · answered May 12 '16 at 16:53

1

I want to get some descriptors for each white area on image to filter that areas and operate with them separately. How can I do it?

My goal could be reached with sklearn.measure.label. This function returns a numpy array with the same shape and labels for each connected area.

But anyway it is still not clear why SimpleBlobDetector from OpenCV doesn't work.

answered May 12 '16 at 16:53

Askhat Nuriddinov

143
1
10

Try skimage blob detection. http://scikit-image.org/docs/dev/auto_examples/features_detection/plot_blob.html. – Chandra Oct 25 '16 at 13:12

score 0 · Answer 2 · answered Jan 07 '17 at 01:22

0

I was having the same problem. I had to remove the fiterbyarea parameter:

params.filterByArea = False

answered Jan 07 '17 at 01:22

Guido

634
1
8
19

Detecting blobs on image with OpenCV

2 Answers2