针对传统主动轮廓模型在目标强边缘处容易产生振荡和弱边缘处容易泄露的缺点,提出了一种基于区域能量最小和主动轮廓模型的医学目标提取模型.这一基于目标灰度统计概率和水平集的主动轮廓分割模型,把能量函数表示为在目标区域内对像素点属于目标概率的积分,并在水平集框架下对能量函数最小化,得到分割的迭代方程;同时,通过附加的速度约束项,使得主动轮廓在越过目标边缘时降低速度,大大提高了分割的收敛性和准确度.通过大量冠状动脉和二尖瓣的分割实验以及与几种传统主动轮廓模型和手工提取的比较,表明该模型在医学图像分割方面的健壮性、准确性和有效性.%In order to solve the problems of traditional active contour models, which move in a high speed on strong edge or leak out on weak edge, this paper proposed a medical object extraction mode) which was based on regional energy minimization and active contour model. The model employed objects' statistical intensity distribution in a level set framework, and ex-pressed energy function as an integral of probability of pixels belonging to objects. The energy function was minimized in a level set framework which led to a iterating equation. At the same time, an edge based speed constrain term was able to slow down the active contours when they steped over a steep boundary of the objects, which made the extraction procedure more convergent and accurately. As shown in the experiments of coronary and mitral valve extraction with comparison with several classical mod-els and manual outline,the proposed model is able to extraction medical objects in an automatic way and the results are also more robust,accurate,and convergent than several traditional modelg.
展开▼
