A Hybrid Segmentation Approach Based on Fuzzy Connectedness and Voronoi Diagram

This paper presents a hybrid segmentation approach for medical images that requires minimal manual initialization by integrating the fuzzy connectedness and Voronoi diagram classification algorithms. We start with a fuzzy connectedness filter to generate a sample of tissue from a region to be segmented and obtain image statistics that constitute the homogeneity operator to be used in the next stage. The output of the fuzzy connectedness filter is used as a prior to the Voronoi diagram classification filter. This filter performs iterative subdivision and classification of the segmented image resulting in an estimation of the boundary. The output of this filter is a binary image that can be used to display the 2D or 3D result of the segmentation. Comparing with other medical images segmentation approaches, this hybrid approach integrates boundary-based and region-based segmentation methods that amplify the strength but reduce the weakness of both techniques. The collaboration between two different methodologies tends to result in robustness and higher segmentation quality. We have already realized this approach in our medical images application and got a satisfying result.