中文摘要：

为了消除主动轮廓模型耗时的重新初始化过程,在分析了主动轮廓曲线演化特性的基础上,引入改进的符号距离函数的约束项到变分GAC模型中,该约束项是一个非线性热方程,通过对非线性热方程扩散率的归一化,使水平集函数始终快速稳定的保持符号距离函数的特性,新算法减少了迭代次数和运行时间。另外,新算法采用灵活的初始化方法和保持曲线二维梯度和散度算子旋转不变性的离散化数值方案,提高了分割算法对模糊边缘的辨别能力。实验表明该方法是有效的,对噪声有很强的鲁棒性。

英文摘要：

In order to eliminate the time-consuming re-initialization procedure of active contour model,an improved restriction item for signed distance function is introduced into variational GAC model on the basis of analyzing the evolution characteristics of active contour curve.To maintain the properties of the signed distance function quickly and stably,the level set function is supplemented with the proposed restriction item that is a nonlinear heat equation with normalized diffusion rate.The novel algorithm reduces the number of iterations and running time.In addition,a flexible initialization method and more efficient discretization method with spatial rotation-invariance gradient and divergence operators are proposed as the numerical implementation scheme to improve the ability to distinguish the fuzzy edges.Experiment results show the effectiveness and strong robustness against noise of the proposed algorithm.