基于先验形状的CV模型肝脏CT图像分割  

Liver CT image segmentation based on prior shape CV model

作　　者：王小芳[1] 赵于前[1] 

机构地区：[1]中南大学信息物理工程学院,湖南长沙410083

出　　处：《光电子．激光》2010年第6期953-956,共4页Journal of Optoelectronics·Laser

基　　金：中国博士后科学基金特别资助项目(200902482);湖南省自然科学基金资助项目(09JJ3119);湖南省博士后科研资助计划资助项目(2008RS4026);湖南省科技计划资助项目(2009FJ3015);中南大学研究生学位论文创新基金(2009SSXT177);中南大学大学生创新创业启航行动项目(2009CX23)

摘　　要：针对目标被部分遮挡或部分信息丢失情况下CV模型不能正确识别的问题,提出一种新的分割算法。首先,利用数学形态学对原肝脏图像进行滤波,并结合其他算法建立肝脏先验形状;然后,采用边缘查找和区域标定等算法,对肝脏先验性状的边缘以及边缘内外区域进行赋值,构建执行效率高的符号函数距离函数,将其通过形状比较函数嵌入到CV模型的能量泛函中,形成新的基于先验形状的CV模型,并将此模型用于分割存在干扰或者被部分遮挡的肝脏CT图像。与CV模型分割结果相比,本文算法能在目标周围存在干扰信息或者被部分遮挡的情形下,成功地正确识别出目标区域。CV model can not segment correctly under the condition that some essential information is missed partly or some parts of the objects are occluded.To solve this problem,a new model based on prior shape focusing on detecting occluded objects is proposed.The new model firstly constructs prior shape which is obtained by mathematical morphology combining with other algorithms,then integrates the prior shape into CV model functional through a novel signed distance function,in which the signed function is constructed by fast implemented boundary searching algorithm and region labeling.The proposed method is applied to segment liver from CT images,with noises around boundary or part of liver being occluded.Compared with the results of CV model,the experimental results show that the new model can detect occluded liver from CT images successfully.

关 键 词：图像分割 水平集 CV模型 先验形状 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]