Path Integral Based Contour Extraction Method Using Multilevel Metropolis Sampling

摘要

In this paper we proposed a novel path integral method using multilevel Metropolis sampling to extract the contours of interested objects of medical images, which is a quantum statistical approach inspired by the essential characteristics of quantum. The implemented multilevel Metropolis sampling is a fast approach to the sampling problem and can be used to sample any distribution function and calculate all actions for the movements of quantum particle towards the position with high probability density likely to be accepted. Comparing with traditional deformable models based on classical mechanics which are sensitive to the initialization of contour and difficult to adapt topology changes of the contour, our work has the capability to break through the limitations of deformable models and makes the boundary extraction in high noise and low contrast of medical images possible. Experiments were performed with synthetic and medical images and the feasibility and potential of our method was demonstrated.