level sets segmentation problem [3 Attachments]

Started by Esmeralda Ruiz Pujadas September 29, 2011
Hi I am new in this forum so first I would like first to introduce me. I am
in my first
year of my phD and I have a lot to learn yet.

My project is related to segment the
myocardium and I wanted to try with level sets to see how it works. I have
implemented a paper and
It fairly works not so perfect as I thought but it seems a good basis to
keep working on .

I attach some results of my own implementation of that paper. It does not
look bad but still a lot to improve.

I would like to ask you if you could know why it can appear this noise
(output0.tif)
like little curves when the region looks homogenius and if it would be worth
it
to force not to change topology and any idea how to implement that.

Another problem I found in level sets is the forces that sometimes they are
going in opposite directions and if I put low one force and increase another

one the countour may separate from the contour I wish, so in your
expertise do you use something to avoid that??. I try with some weights
using
the gradient multiplied by the shape prior energy but without luck.

I would like to reduce the force from the points where the gradient is
higuer
and increase it in not convex points. If someone could give me an idea
it would be nice. Thanks

Sincerely
Esmeralda Ruiz