[Insight-users] Label image using confidence connection without seeds

[O'Connor, Aonghus]

Hi all,

 

Is it possible to automatically segment a noisy image without defining seeds
so that each segment has a different label?

I know there is a filter that labels a binary image:
ConnectedComponentImageFilter.

 

For example, suppose I have an image filled with gaussian noise of a
specific mean and variance. Within the image are two distinct noisey
elipptical regions of different mean and variance. Without defining seed
points is there a way to uniquely label each region i.e. background: 0,
ellipse1: 1, ellipse2: 2?

 

Maybe theres a way to binarise all of a noisy image by confidence connection
without defining seeds?

 

Any help appreciated.

 

Aonghus.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20070222/c63e16f3/attachment.html