Fast initialization of level set method and an improvement to chan-vese model

The initialization is an important step in the level set method. However it is a computational time-consuming step. In order to speed up the level set evolution procedures, we first introduce a new initialization algorithm based on the vector distance transform, which propagates a vector with a position of the nearest object pixel instead of the scalar distance. A new sign map labeling method, based on the flood fill, is proposed to distinguish the inside and outside of the 2D closed active contour The active contour model proposed by Chan and Vese [1] can detect object whose boundaries are not necessarily defined by gradient. Our additional contribution to this paper is to present a further improvement to C-V model by replacing delta(phi) with \delphi\ to gain the global optimization. Finally, we illustrate the efficiency and performance of the proposed model by experimental results.