Segmentation of Ultrasound Image Based on Cluster Ensemble

Image segmentation plays an important role in both qualitative and quantitative analysis of medical ultrasound images. Recently cluster ensemble techniques have been shown to be effective in image segmentation, but selecting ensemble approaches to combine multiple clusterers is critical problem in image segmentation with cluster ensemble. In this paper a new ensemble approach with the spectral graph theory is proposed. Specifically, base clusterers are obtained by K-means cluster algorithm firstly. Secondly, the similarities matrix is constructed based on results of base clusterers. Thirdly, the image is segmented using cluster ensemble approach which integrates K-means clusters using improved spectral cluster algorithm based on the similarities matrix. Experimental results show that the proposed method performs better than some existing cluster ensemble techniques without high computational cost.