使用行扫描法提取图像轮廓  

Using Row-scanning Method to Extract Image Counters

作　　者：郭鑫[1] 刘浩[1] 

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《机械制造与自动化》2012年第5期104-109,共6页Machine Building & Automation

摘　　要：传统的轮廓提取方法,例如边缘检测、水平集、活动轮廓模型等,存在着对初始轮廓敏感、噪声敏感、难以分割弱边缘和复杂形状轮廓等不足之处。以CT影像中髋关节的轮廓提取为应用背景,提出了一种基于行扫描的轮廓线提取方法。首先,对原始图像进行阈值分割和区域增长,分离出目标区域。然后,对目标区域中的孔洞进行自动填充;在填充过程中,通过目标区域的凸包来判断填充的单连通区域是否在目标区域内。最后,使用逐行扫描的方法搜寻像素信息突变点,得到目标区域的轮廓线。大量的数值实验表明,此算法的运行效率高,对于狭长、分叉、噪声干扰以及存在模糊边界的目标区域都能很好地提取轮廓线。Many deficiencies, such as sensitivity to initial contour and noise, difficulties for segmentation of regions with complex shapes and weak edge exist in the traditional contour extraction methods, such as edge detection, the level set, active contour models etc. Taking CT hip contour extraction of CT images as an application background, this paper presents a contour extraction method based on the row-scanning. Firstly, the target area in the original image is separated out by using the threshold segmentation method and the region growth method. Secondly, the holes in the target area are filled outomatically. In the filling pro- cessing, the filling simply connected region is judged to be in the target region by using the convex hull of in the target region. Einaily, the row-scanning method is used to search the mutation of pixel information, we can get the target area contour. Many nu- merical experiments show that the algorithm is efficient. Furthermore, the algorithm can be also used to the extract contour very well for long and narrow, bifurcate, noise and blurring boundary of target region.

关 键 词：行扫描 轮廓 图像分割 凸包 髋关节 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]