Image segmentation based on Fuzzy hypergraph Model

This paper proposes a fuzzy hypergraph model (FHM) for image representation. This model uses both local and globe information of images rely on location space and color space respectively. And by using hypergraph, this model could represent the complex relationships between more than two pixels instead of simple graph which only represent pairwise relationships. This paper uses a semi-supervised learning method which aims to FHM to segment image. Experimental results on various images show that the proposed method is good at classifying the unlabeled pixels.