基于改进的变分水平集和区域生长的图像分割方法的研究

Image Segmentation Method Research Based on Improved Variational Level Set and Region Growth

针对水平集和区域生长方法都存在对噪声和初始边界敏感以及容易从弱边缘处泄露等不稳定的问题,提出了结合待分割目标灰度统计信息和图像梯度信息的水平集演化函数对水平集方法进行改进,并利用区域生长方法解决水平集方法对初始边界敏感的问题.分别用传统区域生长方法、阈值方法、GAC模型、C-V模型、Snake模型以及本文方法进行从腹部CT图像分割肝脏区域的实验比较,实验结果表明:本文方法不仅可以减少图像分割的时间,而且显著地提高了分割质量.

To address the instability problems of level set and region growth, for example, they are sensitive to noises and initial boundaries as well as they will easily leak from the weak boundaries, an improved image segmentation method based on level set is proposed. Our model consists of an external energy term that involves the image gray-scale statistical information and gradient information. And we use region growth method to solve the problem that level set method is sensitive to initial boundaries, we contrast our improved method with region growth method, threshold method, GAC model, C-V model, Snake model to segment livers from abdominal CT images. The experiment results show that our method can not only be efficient for image segmentation, but also greatly improve the quality of segmentation.