The aim of this paper is to present a fully
automatic adaptive k-means segmentation algorithm for MR Images in a 3D space
. We model the gray scale values of the 3D image with a
White Gaussian Process
and superimpose a prior model on the region process in the form of
Markov Random Field
. These assumptions require the use of estimators for the parameters of the two processes. This has been carried out using decreasing size windows.
The Hammersley-Clifford theorem allows us to model the region process in term of a Gibbs Distribution. The Gibbs parameter
is estimated using a correlation-based technique. The segmentation is obtained maximizing the a posterior density function using an Iterated Conditional Modes technique.
The proposed algorithm is fully automatic, i.e. all the parameters of the model are estimated within the segmentation process.