为快速、准确地从复杂的图像中提取出感兴趣的目标,基于变分水平集框架,结合梯度矢量流和符号距离函数惩罚泛函的优点,提出一种基于自适应外力的图像分割方法.该方法在自适应外力作用下将位于目标内部、外部,甚至与目标轮廓相交叉的初始轮廓准确地吸引到目标的边缘;符号距离函数惩罚项的引入避免了繁琐、耗时的符号距离函数初始化工作;同时,模型中添加了加权弧长调整项以保证曲线演化的连续性和平滑性.最后将该方法与现有的Li的快速变分法用于合成和实际医学图像的仿真模拟,证明了该方法的可行性和优越性.

In order to rapidly and accurately extract regions of interest from a complicated image,by combining the gradient vector flow and the signed-distance penalized functional,an image segmentation method based on adaptive external force is proposed under the variational level set frame.This method correctly attracts the initial contours inside outside,even across the object to the object boundary,introduces the signed-distance penalizing term to avoid complicated and time-consuming signed-distance re-initialization procedure,and employs the weighted arc length-rectifying term to make the contour to evolve continuously and smoothly during the propagation.The proposed me-thod,together with the existing Li＇s method,is finally used to segment hybrid and real medical images,the results demonstrating its feasibility and superiority.