GrabCut在人体序列切片图像分割中的应用  

Application of GrabCut in Human Serially Sectioned Image Segmentation

作　　者：李嘉刚[1] 李小宁[1] 石杰[2] 庄敏[1] 陈戈[1] 

机构地区：[1]中国海洋大学信息科学与工程学院,山东青岛266100 [2]青岛市市立医院,山东青岛266071

出　　处：《计算机技术与发展》2011年第12期246-249,共4页Computer Technology and Development

基　　金：国家自然科学基金项目(60873170);教育部博士点基金课题(200804230003)

摘　　要：将GrabCut算法应用于人体序列切片图像分割,解决手动分割操作繁琐、效率低等问题。在简要介绍GrabCut算法基础上,选取可视化韩国人体数据集(VKH)中肾脏部位的序列图像,利用该算法对肾脏进行分割。通过Visual C++环境,在自主开发的三维人体肾脏结构虚拟现实软件(VRKidney)中实现了肾脏分割、修改功能、同一幅图像分割多个对象功能等,与手动分割法、边界提取法的比较,验证了该方法具有操作简单、高效的特点。研究表明GrabCut算法操作简单、分割效率高,可以很好完成人体序列切片图像的分割。Applying the GmbCut algorithm in human serially sectioned image segmentation can solve the problems of complicated operation and low efficiency in manual segmentation. Based on the essential principle GrabCut, select serial images of kidney from visible Korean human data set and apply the CrrabCut algorithm to achieve renal segmentation. It＇ s implemented in VRKidney platform which is developed under the conditions of Visual C＋＋, the function of modification and segmenting multiple objects in one image are also finished, and the efficiency of the algorithm is proved by comparing with manual method and bound extraction method. The research shows that GrabCut algorithm is easy to operate, high-efficiency to segmentation and can excellently complete ,segmentation.

关 键 词：GRABCUT 图像分割 韩国人体数据集 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]