改进的双水平集医学图像分割方法  

Improved Segmentation Method for Medical Images Based on Dual Level Sets

作　　者：秦绪佳[1] 郑琴[1] 张素琼[1] 韩军 

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310032 [2]大连现代高技术公司数字核医疗部,辽宁大连116025

出　　处：《小型微型计算机系统》2012年第4期887-890,共4页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61075118)资助;浙江省自然科学基金项目(Y1100880)资助

摘　　要：由于基于简化M_S模型的多相水平集图像分割模型仅仅利用了图像的区域信息,对图像的另一个重要信息(边缘信息)没有有效的利用,同时在分割的过程中需要对水平集函数不断进行重新初始化.为了解决上述模型的不足,本文提出改进的双水平集医学图像分割方法.该方法主要是在基于简化M_S模型的多相水平集图像分割模型的基础上将图像的边界信息项和为避免重新初始化水平集函数的惩罚项加入模型中.实验结果表明,添加了边界信息后的模型能够在边界位置定位更容易,同时改进后的双水平集模型在实现多目标分割时,无需重新初始化水平集函数,减少了计算量,简化了算法实现的复杂度.As a multiphase level set model for image segmentation using the simplified M_S model only use the regional information of image, the other important information （ the edge information of image ） is not effectively used. Meanwhile in the process of image segmentation the level set function needs to be continually re-initialized. To solve the limitation of above model, this paper presents the improved dual level sets of medical image segmentation method. The method which is mainly based on a multiphase level set framework for image segmentation using the simplified MS model, adds the boundary irlformation item of image and the penalty item which avoids re-initializing the level set function to the model. Experimental results show that the model added the boundary informa- tion item can more easily locate the boundary position. Meanwhile, when the improved two-level set of medical image segmentation implements the segmentation of multi-objective, it doesn＇t need to re-initialize the level set functions. This will reduce the computation and simplify the implementation complexity of the algorithm.

关 键 词：医学图像 图像分割 多相水平集 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]