Research on Fast Initialization for the Level Set Method

摘要

A novel composite approach for initializing the level set method is proposed. The initialization is computational time-consuming and crucial for the application of the level set method, such as image segmentation. The composite algorithm is based on the fast marching method and the implicit iteration method, and moreover, the advantages of two methods are respectively incorporated in our algorithm. Firstly a simple labeling approach based on the fast marching method is developed in this algorithm. The signature function is suitably modified to solve the numerical stability problems when the implicit iterative method is used to the construction of the signed distance function. At last, comparison with the previous algorithms shows that the proposed approach is accurate, simple and efficient.