摘要：

针对CV模型的多水平集算法需要较高的数值稳定性以及曲线演化速度慢的缺点，根据人脑MR图像的特征，提出一种快速CV双水平集算法，统计被2条曲线划分成4类的直方图，构造符号矩阵，依次将直方图上的点放入其他类中，根据能量的变化更改该点对应点的符号，得到粗分割结果，并对粗分割结果进行优化。对MR图像进行的分割实验表明，其分割效果更好，速度有大幅度的提高。

关键词: CV模型, 直方图, 图像分割

Abstract: Multiphase level set method of Chan-Vese(CV) model is not suitable for real-time application for having numerical stability constraints and its low efficiency. Aiming at this disability and based on the specialty of human brain, a fast method to solve is developed. This approach computes the histograms of four areas which can be used to distinguish the area cut by two curves, and constructs signed tables. The points of the histogram are changed to other clusters and the sign of the points are checked according to the change of the energy. And the improved mean of small neighborhood method is used to optimize the results and get the final edges. Experimental results show that the new model can get the better results in an efficient way.

Key words: Chan-Vese(CV) model, histogram, image segmentation

中图分类号: 

• TP391.41