Segmentation of medical images using a mixture model and morphological filtering

In this work a practical solution for segmenting 3D MR images is proposed. This method is based on a mixture model and Expectation Maximization (EM) algorithm. Here, we only focus on image segmentation which is used as a first step in our 3D compression and visualization system we are developing. A pretreatment based on gray-level thresholding followed by a morphological filtering is employed, then a stochastic segmentation method based on a finite mixture model is used. The obtained results confirm that statistical segmentation based on mixture model combined with Bayesian decision rule is a powerful tool for segmenting MR images.