基于区域混合活动轮廓模型的医学图像分割

Medical Image Segmentation Based on Region-based Hybrid Active Contour Model

针对变分水平集算法在图像分割过程中计算量较大且收敛速度慢的现象,在一些基于区域的活动轮廓模型基础上提出了一种新的基于区域混合模型的非凸正则化活动轮廓模型。该模型构造了一个新的能量泛函,该能量泛函结合了考虑图像局部聚类性质的LBF模型和测地线模型,增加了非凸正则化项,加快了轮廓曲线的收敛速度,可以很好地保持区域形状并能防止边缘过平滑,然后通过经典有限差分法求得能量泛函的极小值。最后,在合成图像和医学图像上做了仿真实验,结果表明,该算法具有较快的收敛速度和很好的鲁棒性,分割结果也较准确。

In view of the phenomenon that the calculation of variational level set algorithm is much larger and the speed is too low in the process of image segmentation, this paper proposed a new region-based hybrid nonconvex regularization active contour model based on some region-based active contour models. This model constructs a new energy functional which incorporates the LBF model having the property of local clustering of an image and geodesic active contour mo- del. By adding a nonconvex regularization term, they fasten the convergence speed of the contour curve, and can well pre- serve the shape of the region and protect the edge from oversmoothing. Thus, the minimum of the energy functional will be obtained by the typical finite difference method. Results of the simulation experiment on synthetic and medical images show that the proposed algorithm has quite fast convergence rate, accurate segmentation results and better robustness.