Segmentation of biomedical images and volumes
Gil Gaillard
Section Microtechnique
Student Project
Semester Project, February 2000
Abstract
The goal of this semester project was to find an algorithm of segmentation of biomedical images and volumes. These have to be segmented in two labels only (background and foreground).
The Markov random fields are used to establish a model of the image where the random variables are the labels. These labels are hidden that's why this model is called Hidden Markov Model (HMM). This model is useful to find the labels of each pixel considering its value and the pixels around it.
The Maximum likelihood method allow us to find the labels explaining the values the best with the HMM.
Result 2D
image result
Result 3D
volume
slice
result
slice
