Markov chain

v s ges re d ithm dat esp results on synthetic and real data demonstrate that Markov surfaces can segment regions that are defined s in m ology,. Sever s not seems necessary in these cases. Several methods [1–5] have been proposed for semiautomatic segmentation. A live wire algorithm is proposed in [1] to formulate boundary extraction as a graph searching problem. It utilizes the start points and the end points specified by users, and generates paths between these points using local gradient features. Falcao trol during the curve evolution. Ardon et al. [3] proposed a surface extraction method based on the start and end curves specified by users. A network of minimal paths between these curves are generated using Fast Marching method, and a 3D surface is acquired by the interpolation between minimal paths. A 3D level set algorithm is performed for segmen-tation using the acquired 3D surface as initialization. Although this approach may provide good results, the topology of the network is often problematic [4]. An implicit method is proposed in [4] for this issue. It segments the object by finding a 3D real function using ⇑ Corresponding author