基于改进的Snake模型的脑部MR图像分割方法*

为实现对脑部磁共振图像的分割,提出了一种改进Snake模型的图像分割方法。通过引入轮廓中心的概念,在贪婪Snake模型的能量函数中增加距离势能作为外部约束能量,增大了外能的吸引范围,使分割结果不依赖于初始轮廓;对各能量项进行归一化操作,并以归一化扩散方程各分量的梯度矢量流代替MR图像的梯度,提高了模型处理弱边界和深度凹陷区域的能力;对各能量函数的离散化和参数的选择进行了阐述。实验结果表明,该算法是一种有效的分割脑部MR图像的方法。

Brain MR image segmentation based on improved Snake model

To segment magnetic resonance brain images,a new method based on an improved Snake model was proposed.The method by introducing contour center,added a distance potential energy term as exterior sanction energy to the energy function of greedy algorithm for capturing the image feature in a wider range,so the result of segmentation didn't rely on the initial contour.By normalizing every energy term and replacing the gradient field by improved GVF with normalized force vectors in the diffusion equations,the model could segment the weak edges and deep concave regions effectively.Elaborated the discretization of energy function and choosing of parameters.The experiment results demonstrate that this algorithm is an effective method for segmenting the brain MR image.