A Cooperative Method to Improve Segmentation of Brain MR Images

In this paper , we present a fully unsupervised segmentation process of magnetic resonance image (MRI) of the brain using a data fusion technique and some of ideas of the possibility theory context. The fusion methodology is decomposed into three fundamental phases. We modeling information coming from T2 and PD weighted images in a common framework, in this step an hybridization between FCM and PCM algorithms is retained. In the second phase an operator of fusion is used to combine then these information. Fi nally, an image of fusion is generated when a decision rule is applied. Some results are presented and discussed using a set of simulated MR image.