[Insight-users] Re: Splitting algorithm

[Sayan Pathak]

Hi SriValli,
One way to look into the problem of splitting is to use the =
classification framework.=20
You could customize a decision rule using the base class
http://www . itk . org/Doxygen12/html/classitk_1_1DecisionRuleBase . html
and tailor it such that it meets the splitting rule you would like to =
have. Plug-in the decision rule into the classification framework and do =
the splitting of a clustered object into two or more classes.=20

You could also look into the k-means model estimator where you can =
specify the number of clusters you want and the estimator will compute =
the region statistics that will partition the cluster into the number of =
user specified clusters. However, this is an intensity based =
partitioning. You would have to come up with your own version if you are =
interested in shape based partitioning of a given cluster.

The classes that presently use the classifier framework are documented =
under the following link:
http://www . itk . org/Doxygen12/html/group__ClassificationFilters . html

Thanks,
Sayan




--__--__--

Message: 7
From: "cspl" <affable at hd2 . dot . net . in>
To: <insight-users at itk . org>
Date: Fri, 27 Jun 2003 09:52:30 +0530
Subject: [Insight-users] Splitting Algorithm

This is a multi-part message in MIME format.

------=3D_NextPart_000_0014_01C33C91.D3257E00
Content-Type: text/plain;
	charset=3D"iso-8859-1"
Content-Transfer-Encoding: quoted-printable




Hi all !

Our requirement is to split clustered objects using convex split =3D
algorithm. Can anyone suggest us on usage of ITK class for the =3D
requirement.

Regards,
SriValli