基于Chan-Vese模型的医学图像分割算法  

Medical image segmentation algorithm based on Chan-Vese model

作　　者：职占江[1] 宋锦萍[1] 

机构地区：[1]河南大学数学与信息科学学院

出　　处：《计算机工程与应用》2008年第14期190-194,共5页Computer Engineering and Applications

基　　金：国家自然科学委员会与中国工程物理研究院联合基金(No.10576013);河南省自然科学基金(the Natural Science Foundation of Henan Province of China under Grant No.0611053200)

摘　　要：在二维Otsu方法和Chan-Vese模型的基础上,提出了一种新的医学图像分割算法。首先用二维Otsu方法将图像分成目标、背景、边缘和噪声等4部分,然后通过一个能量函数来判断边缘和噪声区域中各像素点属于背景还是目标,并利用同质区域的全局信息对初始分割结果进行微调,得到更精确的分割效果。该算法优化了初始轮廓位置,有效地解决了初始位置影响曲线演化速度问题,并通过逐点代入法来极小化能量函数,减少了计算量,提高了图像分割的速度。实验结果表明,提出的算法具有抗噪性,分割效果良好,有很好的实际意义。This paper presents an OCV medical image segmentation algorithm based on two-dimensional Otsu method and Chan-Vese model. First,we use two-dimensional Otsu method segment the image into the objectives,background, marginalized and noise of four parts then through an energy function to judge edges and noise pixels belonging to the background or target, and the use of regional homogeneity of the global information initial segmentation results tuning, to be more precise segmentation. The algorithm optimizes the initial profile position to address effectively the impact of the initial location of the evolution curve speed, by through point by point to the generation of energy function for the minimal value ,to reduce the amount of computation ,improve the speed of image segmentation. Experimental results show that the proposed algorithm can resist noise,the result is good,a very good practical significance.

关 键 词：医学图像分割 二维OTSU方法 水平集 CHAN-VESE模型 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]