融合边缘和区域信息的水平集矢量图像分割

Level set valued-vector image segmentation with combination of boundary and region information

在充分利用矢量图像各通道区域和边缘信息的基础上,变分IAC(集成活动轮廓)模型引入了非线性热方程的符号距离函数约束项,使水平集不用耗时的重新初始化而始终保持符号距离函数的特性。对非线性热方程传导率的均衡化,使水平集的演化分割过程快速稳定。另外,算法改进了曲线2维梯度和散度算子传统离散化方式,使梯度和散度算子保持空间旋转不变性。实验结果表明,该方法是有效的,提高了分割的准确性和鲁棒性。

A restriction item that is a nonlinear heat equation is attached to a variational IAC （integrated active contour） model on the basis of analysis on regions and edges information from all channels of the valued-vector image, which forces level set to maintain the signed distance function properties without the costly re-initialization. A balance function for diffusion rate of the nonlinear heat equation is introduced into this model, and therefore the level set evolution segmentation process becomes fast and stable. In addition, an efficient discretization method with spatial rotation-invariance gradient and divergence operator are proposed as numerical implementation scheme. Finally, the experiments on some images have demonstrated the efficiency, accuracy and robustness of the proposed method.