A fuzzy clustering image segmentation algorithm based on Hidden Markov Random Field models and Voronoi Tessellation

- Using VT and HMRF model to improve robustness and noise insensitiveness.
- Combing VT-HMRF model into the FCM based framework.
- Comparing results with traditional FCM based algorithms.

In this paper, we present new results related to the Voronoi Tessellation (VT) and Hidden Markov Random Field (HMRF) based Fuzzy C-Means (FCM) algorithm (VTHMRF-FCM) for texture image segmentation. In the VTHMRF-FCM algorithm, a VTHMRF model is established by using VT to partition an image domain into sub-regions (Voronoi polygons) and HMRF to describe the relationship of neighbor sub-regions. Based on the VTHMRF model, the objective function of VTHMRF-FCM is defined by adding a regularization term of Kullback-Leibler (KL) divergence information to FCM objective function. The proposed algorithm combines the benefits stemming from robust regional HMRF and FCM based clustering segmentation. Segmentation experiments on synthetic and real images by the proposed and other improved FCM algorithms are performed. Their results demonstrate that the proposed algorithm can obtain much better segmentation results than other FCM based methods.