Dear Evren,

I would go with a border detection filter, such as Sobel or maybe Gabor (see links attached).

Hope this helps,

Fernando

https://en.wikipedia.org/wiki/Sobel_operator

https://en.wikipedia.org/wiki/Gabor_filter

Dear Evren,

I would go with a border detection filter, such as Sobel or maybe Gabor (see links attached).

Hope this helps,

Fernando

https://en.wikipedia.org/wiki/Sobel_operator

https://en.wikipedia.org/wiki/Gabor_filter

In another of your threads, I suggested to homogenize the image first.

Can you post an original image? This one has too large jpeg artifacts (see attachment); edge-detection will thus produce many non-real edges.

BTW are you sure the vessels end there? I found a similar image in another thread, showing vessels crossing that 'border'!

yes, my images are taken from prematurity babes(Retinopaty of Premature disease)

your images maybe are diabetic disease. My images regions are very little different color so ı dont find this region.

OK.

And..., do you have better quality images?

Dear Evren,

You can use to Laplacian or Canny  edge detection algorithm.

You can try retinal vessel tracking algorithms based on Gabor wavelets ,Oriented wavelets ,Orientation scores. For example

http://bmia.bmt.tue.nl/people/RDuits/bekkersduitsARCHIV.pdf

Or follow other techniques to model the retinal vessels:

http://digital.liby.waikato.ac.nz/conferences/ivcnz07/papers/ivcnz07-paper24.pdf

Do edge detection.

Dear Erven,

I would suggest you to try the Multi-scale Frangi vesselness filter.

http://www.tecn.upf.es/~afrangi/articles/miccai1998.pdf

In the following link you can find a nice-easy to use matlab implementation of this filter

http://www.mathworks.com/matlabcentral/fileexchange/24409-hessian-based-frangi-vesselness-filter

Hope this helps and good luck

Hi Mohammed, I could not get it working for this image (I converted to b&w and inverted the image).

Probably, the 2nd order derivatives in the Hessian matrix were too small, or the jpeg image suffers too much from compression artifacts.

Anyway, thanks for pointing out that matlab code; it works quite well for angio X-rays!

Hi Evren

I published many works about this subject as below

paper.ijcsns.org/07_book/200911/20091134.pdf

www.aessweb.com/pdf-files/Jasr-pp-307-309.pdf

Regards

I know this an old thread and am not sure if you still need assistance, but I would defintely recommend you visit www.MIPAR.us.

It is very powerful and intuitive 2D/3D image analysis software written by scientists/end-users and is about to be released as a free trial on December 14th. You can sign up on the site as well as submit images/datasets to test.

There are many image analysis options available, but I think you may just find MIPAR to be something special if you give it a try!

Dear John,

Could you please stop spamming this forum with advertising commercial software. Go ahead if you have some segmentation result or if you have something to say about algorithms incorporated in MIPAR that would be able to tackle this problem...

This is a scientific forum.

 Dear Lambert,

First, please let me apologize for any irritation I may have caused you, with what you call spamming. I sensed a great deal of frustration in your post, and you instilled a great deal in me. Much as I would like, I do not have the time to post a solution to every poster's segmentation problem, although I have for quite a few. Therefore, for the others, I have decided to "advertise" MIPAR to as many members as possible, knowing with great confidence it has the potential to help them, as it has myself and many others.

Contrary to your reaction, I've received many positive responses, and, when those responses included images, have return those images processed, to the gratitude of many community members. I see dozens of repetitive posts of "advertising" freeware solutions like ImageJ, without a single result or follow-up, and strangely I never see your public complaints on those threads. Believe it or not, commercial products and a desire to help others are not mutually exclusive ideas - and MIPAR is currently free. Anyway, I don't want to irritate other members any further, many of whom have been very appreciative of my postings, and have had no desire for your policing. Here is a potential solution to Evren's problem. A masked adaptive threshold, followed by classic morphology and size filter steps. Evren, feel free to contact with any questions, and I'd be happy to share the MIPAR recipe I used if you receive a free copy.

I will continue to positively contribute to this scientific forum.

Dear John,

I agree with you that answers to questions like "use MatLab" or "use ImageJ" are also not very helpful. I probably did not complain about those, because such answers are made by nitwits. I got the impression that you  are more knowledgeable in image-processing and therefore worth my criticism. Also I had the impression that MIPAR is commercial software, because you wrote about free trial-versions.

Anyway, I appreciate your scientific contributions, but if you do not have the time to post solutions to problems or suggestions towards solutions, then I would say: don't post an answer.