# canny segmentation level set
**URL:** https://discourse.itk.org/t/canny-segmentation-level-set/1896
**Category:** Algorithms
**Created:** [May 20, 2019, 10:17am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896 "2019-05-20T10:17:11Z")
**Posts on this page:** 8
**Page:** 1
### Author:  [@Marco\_Festugato](https://discourse.itk.org/u/Marco_Festugato)
#### Post date: [May 20, 2019, 10:17am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/1 "2019-05-20T10:17:11Z")
hi guys, thank u in advance for reading this.
Can I use the Canny Edge level set to refine this segmentation I obtained using a simple binary thresholding? I want to delete all those little points and keep only the big ones (they are vessels)…
---
### Author:  [@mattias](https://discourse.itk.org/u/mattias)
#### Post date: [May 20, 2019, 10:45am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/2 "2019-05-20T10:45:45Z")
I would use a binary eroding filter until only remnants of the large regions remain, then use that map as seeds to reconstruct the original regions.
---
### Author:  [@Marco\_Festugato](https://discourse.itk.org/u/Marco_Festugato)
#### Post date: [May 20, 2019, 11:30am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/3 "2019-05-20T11:30:30Z")
Hi @mattias and thank u for your reply! Do u know if I can use that level set even if the regions I want to segment are not ‘‘closed’’ regions?
Thanks in advance for any help
---
### Author:  [@mattias](https://discourse.itk.org/u/mattias)
#### Post date: [May 20, 2019, 11:47am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/4 "2019-05-20T11:47:45Z")
This is the result I get if I first label all connected regions then keep the four largest regions in your image.
So I think that’s a possible solution for you as well.
---
### Author:  [@Marco\_Festugato](https://discourse.itk.org/u/Marco_Festugato)
#### Post date: [May 20, 2019, 11:50am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/5 "2019-05-20T11:50:38Z")
Thank u again Mattias for your quick answer. Can I ask u which values of the level set parameters did u use?
---
### Author:  [@mattias](https://discourse.itk.org/u/mattias)
#### Post date: [May 20, 2019, 11:55am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/6 "2019-05-20T11:55:34Z")
Since I don’t have the original data I simply used the equivalent of a Connected Component Image Filter to label each connected region in your binary image then I used a Label Shape Statistics Image Filter to extract the four largest regions.
---
### Author:  [@matt.mccormick](https://discourse.itk.org/u/matt.mccormick)
#### Post date: [May 21, 2019, 4:30am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/7 "2019-05-21T04:30:02Z")
Another good option to be aware of is the [BinaryOpeningByReconstructionImageFilter](https://itk.org/Doxygen/html/classitk_1_1BinaryOpeningByReconstructionImageFilter.html).
---
### Author:  [@Marco\_Festugato](https://discourse.itk.org/u/Marco_Festugato)
#### Post date: [May 21, 2019, 6:20am UTC](https://discourse.itk.org/t/canny-segmentation-level-set/1896/8 "2019-05-21T06:20:32Z")
thank u both for your reply!