具有部分多目标轮廓提取特性的单水平集方法

C-V模型只能实现单目标图像的轮廓提取，造成这种后果的主要原因是C-V模型过多地考虑了演化曲线内外部的信息，反而造成了曲线的不正确演化。针对这点，提出了一种新的更为简单的水平集方法，该方法对初始曲线位置选取没有特殊要求，演化过程中符号距离函数也无需重新初始化。运用该方法，给出了几幅不同性质图像的实验结果，结果表明，方法不仅能保证单目标轮廓的正确提取，还具有部分多目标轮廓提取的特点，而且需时少，具有一定的抗噪性。

Partial multi-objects extracted based on single level set

The Chan and Vese mode can only extract single object contour,because it considers too much of the interior and exterior information.That leads the evolving curve to a wrong position for the multi-objects image.Focusing on the limitation,this paper proposes a new simple mode,which only considers the exterior information of the evolving curve,and presents several experimental results under different conditions.It shows that this mode not only keeps the CV mode＇s advantages,but also can implement contour extraction on a part of multi-objects images.Moreover,the initial curve can be anywhere in the image,and the sign distance function doesn＇t need to be reinitialized.Meanwhile,it is efficient and antinoise.