[ITK-users] Understanding the Geodesic Active Contour

[bipul]

Hi,

I was going through the implementation for the Geodesic Active Contour
filter and have the following questions about it:

1. The propagation term of the general level set is the speed image in case
of the geodesic filter, which is also the feature image, and in the given
example it is the output of the sigmoid filter. The advection term is the
gradient of the same speed image. I wanted to know the reason for using the
speed image for propagation(since it is supposed to be a constant) and also
for advection. How do these forces cancel each other out when a boundary is
detected? I assume, they must be tuned first for best boundary detection.
Also, what might be other choices for the speed image?

2. Another input to the filter is an initial level set, which is a zero
level set provided by the output of the fast marching filter. I could not
find the use of this input in the implementation. Could you point me to the
class where I can find the implementation that this level set is modified
with each iteration? Also, I guess, when the contour evolves, we only do the
calculations at the boundary of the contour and not the entire volume?

I tried looking up for these questions on the forum, but didn't find any
answers. I would really appreciate it if you could answer these.

Bipul Tarafdar




--
View this message in context: http://itk-insight-users.2283740.n2.nabble.com/Understanding-the-Geodesic-Active-Contour-tp7590154.html
Sent from the ITK Insight Users mailing list archive at Nabble.com.