Curve/Contour Closing/Completion in Image Segmentation

Hi,

I'm trying to get a segmentated image from a bunch of partially overlaping rock particles to analyse their size and shape distributions.

I already got a proper edge detection result (using a multidimensional canny edge detector), but there are still many not connected curves in the image that represent parts of the outline of a particle. So I need to connect them to get closed curves/contours that represent the actual outline well.

There are some scientific papers available on the internet but none of them explain this part of their Image Segmentation algorithm well. Many of them use Anisotropic Contour Completion ACC (sometimes also called Anisotropic Diffusion, which is quit misleading), but they only give some equations with always the same not sufficient explanations along with it.

So it would be great if you could help me to either find a proper algorithm or give some hints on how to understand ACC.

Thanks, Sebastian W

PS: Here are some links to the papers using some kind of ACC I already read. http://www.springerlink.com/content/4hcf5xdt6jqyyc0n/ http://www.ncbi.nlm.nih.gov/pubmed/21097100 http://www.sciencedirect.com/science/article/pii/S1077314204002140

Answer 1

[Image Analyst]

Here's some MATLAB code for it:

http://www.csse.uwa.edu.au/~pk/Research/MatlabFns/#anisodiff

[Sebastian W]

Thanks for your respond, but that's not what I was searching for. There is some misuse in the term Anisotropic Diffusion, some authors called it Anisotropic Contour Completion which does fit better. Already corrected my original post.

[Image Analyst]

If your searches didn't turn up any MATLAB code, there there probably is none available online. You'll have to write it yourself, or try asking some authors for theirs.

Answer 2

[Javier]

Hi, some news? I'm searching also for the same! Thank you