Prev: use input from GUI to run another .m file, show output in GUI
Next: how to download Matlab trial
From: LFG Tech on 24 Mar 2010 02:18
Hello friends,

I am trying to recognise a pattern in an image.

http://whyfiles.org/coolimages/images/csi/fundus.jpg

The pattern in question are nerves . They meander through the length of the image as shown in link above. it is this pattern that I wish to detect.

My approach was to take an edge image and look for parallel lines(as each nerve has a parallel shape).

Cld you advice me as to how this could be achieved?

regards
From: us on 24 Mar 2010 04:17
"LFG Tech " <vinodkaruvat(a)gmail.com> wrote in message <hocaqv$66a$1(a)fred.mathworks.com>...
> Hello friends,
>
> I am trying to recognise a pattern in an image.
>
> http://whyfiles.org/coolimages/images/csi/fundus.jpg
>
> The pattern in question are nerves .

huh(?)...
nerves(?)...
blood vessels...

us
From: David Young on 24 Mar 2010 04:20
Are you sure you can see nerves in that image? The prominent darker structures are blood vessels, surely?

If you want to find the blood vessels, looking for parallel edges probably isn't a good approach - it's too hard to get reliable edges at both sides of such narrow structures. One simple way to start might be to convolve with a centre-surround mask (e.g. fspecial's 'log' mask), then threshold, then use a morphological operation to find the skeleton.

(The convolution mask's negative centre should have a diameter of roughly the width of the structure's you're after, and make sure that the size of the mask array is large enough not to truncate the mask too much - fspecial doesn't look after this automatically.)

If that doesn't work, there are more sophisticated things you could try - but see how you get on with a straightforward method.
From: LFG Tech on 24 Mar 2010 04:33
"David Young" <d.s.young.notthisbit(a)sussex.ac.uk> wrote in message <hoci06$mk1$1(a)fred.mathworks.com>...
> Are you sure you can see nerves in that image? The prominent darker structures are blood vessels, surely?
>
> If you want to find the blood vessels, looking for parallel edges probably isn't a good approach - it's too hard to get reliable edges at both sides of such narrow structures. One simple way to start might be to convolve with a centre-surround mask (e.g. fspecial's 'log' mask), then threshold, then use a morphological operation to find the skeleton.
>
> (The convolution mask's negative centre should have a diameter of roughly the width of the structure's you're after, and make sure that the size of the mask array is large enough not to truncate the mask too much - fspecial doesn't look after this automatically.)
>
> If that doesn't work, there are more sophisticated things you could try - but see how you get on with a straightforward method.

Hello Sir,

It is the blood vessels(not nerves). The blood vessels can be extracted by using a method called as the Kirch's method. But, my aim is to confirm if the image has the blood vessels. say that i have detected the blood vessels- how to confirm this detection other that visually inspecting the image(this is important for the later stages of the algorithm).



regards,
Lfg
From: us on 24 Mar 2010 04:48
"LFG Tech "
> It is the blood vessels(not nerves). The blood vessels can be extracted by using a method called as the Kirch's method. But, my aim is to confirm if the image has the blood vessels. say that i have detected the blood vessels- how to confirm this detection other that visually inspecting the image(this is important for the later stages of the algorithm).

it depends on the output of your kirch routine...
can you tell more about it...

us
 |  Next  |  Last
Pages: 1 2
Prev: use input from GUI to run another .m file, show output in GUI
Next: how to download Matlab trial