Research on image segmentation model based on local hybrid energy fitting

Active contour models (ACMs) have been a successful method for image segmentation. To well segment the images with intensity inhomogenity and overcome the defect of the results highly depending on the initial position of the contour, we propose a new region-based ACM, which combines Hellinger distance to segment images under the framework of variational level set. Firstly, we utilize Hellinger distance to merge two ACMs. By measuring the distance between the real image and the fitted image in the local region, the similarity between them can be revealed and the pixels can be classified according to the distance. Then, combining with the local bias field of an image to construct a new loyalty term, the variational level set function is used to minimize the functional energy. Finally, the experimental results on synthetic, magnetic resonance (MR) and real images with high intensity inhomogeneity show that the proposed model can obtain better performance than the state-of-the-art ACMs, and take less running time. In addition, the proposed method can be applied to other local fitting-based models to improve the robustness of initial contours.