Prev: ODE OutputFcn
Next: plotting an implicit 3d function
From: geoleo Aravindakshan on 15 May 2010 22:57
Hello,

I have some big coin Images, which I have processed using Image processing techniques, and I have extracted Edges (curved edges of coins), I am observing some breaks in the edges. So I have to apply Edge linking algorithms for such curved edges. How can I do - I am finding methods like Hough transform, Morphological operations - using magnitude and angle etc. But not getting a clear idea. I require a matlab code. My final image have extracted edges is a binary image. with background zero intensity and edges of intensity 1. Thank you.
From: ImageAnalyst on 15 May 2010 23:28
geoleo Aravindakshan:
One way to do it would be to use the convex hull. You could fill all
the blobs (edges) in your binary image, and then label all the blobs
(edges) and then call regionprops. Then look at the perimeter^2/area
ratio and identify all non-round blobs. Then for each non-round blob,
find the closest other non-round blob and get the convex hull of this
pair. That will connect them together. This should work. There are
a few sub-steps there that I didn't fully explain but if you think
about it, it will be fairly obvious and straight forward what to do.
There are a few other things you could do if you really really need to
be more accurate but I don't think that should be necessary.

Another way is to try to close the gaps with a morphological closing.
This will work only for gaps smaller than the kernel window, and not
in general for any gap size like the previous method I outlined.

Next time post your images before you ask questions about them - it
will get you better answers.

By the way, have you seen my demo on coin segmentation here?
http://www.mathworks.com/matlabcentral/fileexchange/25157

Good luck,
ImageAnalyst
 | 
Pages: 1
Prev: ODE OutputFcn
Next: plotting an implicit 3d function