Hello Everyone, can Anyone tell me which segmentation techniques I follow i applied ostu thresholding instead of this what other techniques can i use. and for splitting of two circles I used watershed Techniques
I would be very grateful if anyone help me out Thanks
This type of problem is known as blob detection.
For the sharp carbonate you have, you are fine with the watershed transform or the Hough transform. These techniques can handle the overlaps occurring in some parts well.
However, depending on your application, the amount of overlapping as well as the appearance of other artifacts in your slide my complicate things.
If this is only the best possible scenario and/or the configuration will evolve as time goes by (which is a realistic assumption), then you may need some preprocessing.
Hi Prachi,
If all your images are like this one, then you used proper techniques. Did you get good results from Otsu and watershed?
You may consider to normalize the images and then send them to a segmentation model.
You can also try other unsupervised approaches like k-means:
https://ch.mathworks.com/help/images/ref/imsegkmeans.html
As your images do not seem to be too complicated to be segmented, most probably complicated algorithms such as deep-learning-based segmentation models are not required.
All the best,
Amirreza
This type of problem is known as blob detection.
For the sharp carbonate you have, you are fine with the watershed transform or the Hough transform. These techniques can handle the overlaps occurring in some parts well.
However, depending on your application, the amount of overlapping as well as the appearance of other artifacts in your slide my complicate things.
If this is only the best possible scenario and/or the configuration will evolve as time goes by (which is a realistic assumption), then you may need some preprocessing.
Hi, since the features are quite sharp and well defined, I assume a template matching based approach with a circular template should do the job. For the watershed segmentation which should work as well, you should consider that you might end up in oversegmentation, i.e., a lot of small regions which have to be combined again. So you should have a strategy to overcome this problem, e.g using hierarchical watersheds.
Please, instead of "sharp carbonate" read "sharp image." Sorry for my mistake.
Actually @Christoph Heinzl, for that image in particular, a plain edge detector with any simple clustering algorithm can do it. However, life in general is not like that...
Hi Prachi, you could convert your image to gray scale image, threshold it to obtain a binary image and use the famous Hough transform to detect/separate the discs in your image.
Thanks Langlard , but in this type of images what features i will encounter for accurate detection
Prachi Arora, once you have binarized your image, if you use matlab for example you can call the function imfindcircles(image, radiusRange) where you just need to specify the range of the radius of the circles that your would like to detect. The smaller the range, the better in general. Then, the algorithm will give you the center positions of the detected discs and the corresponding radius.
For more information you can have a look to:
https://fr.mathworks.com/help/images/ref/imfindcircles.html
With OpenCV, you can use the function HoughCircles() where you have more inputs to give (and so more flexibility) but the principle stay the same. You can have a look to:
https://docs.opencv.org/master/dd/d1a/group__imgproc__feature.html#ga47849c3be0d0406ad3ca45db65a25d2d
- Badges
- Science topic
- Similar topics
- Computer Science
- Images