Interactive Image Segmentation Based on Graph Cuts and Automatic Multilevel Thresholding for Brain Images

摘要

This paper describes an iterated automatic histogram multilevel thresholding combining with graph cuts algorithm for image segmentation. The contribution of this work resides in the good performances of segmentation obtained. Our objective is to increase the efficiency tumor segmentation shown in brain images in order to bring visual information for diagnosis help. Image thresholding is very useful tool to separate objects and backgrounds. Standard graph cuts consist to found an optimal solution to a wide class of energy functions. The proposed algorithm start with initial segmentation using multilevel thresholding, the segmented regions, pixels are considered as the nodes in the graph cuts. Experimentally our results are much better than segmentation obtained by graph cuts algorithm, this method can be employed to other types of images as well without the influence of λ well without the influence of λ.