利用动态链接表的二值图像目标区域分割

Target area segmenting using dynamic chained list in binary images

为了准确、快速地对二值图像中的连通区域进行标识,分割出各个目标物体的位置和面积,提出了一种基于动态链接表的连通区域标记方法。以目标像素线段为基本处理单元,令像素扫描和连通标识同步进行,同时,建立所用连通体的动态链接表,仅需一次扫描便可根据链接关系将图像中所有目标物体分割出来。定义"中空"的4邻域模板,通过特定的判断运算对二值图像进行开运算滤波,一次遍历便可去除图像中所有前景和背景噪声。实验证明:动态链接表有效地解决了其他算法中重复标记的问题,对于512×512的图像,算法时间约为3ms,且该算法不受所标记物体的形状和数目影响。满足了目标分割在工程上的速度快、准确度高、鲁棒性好等要求。

In order to label the connected components fast and exactly in binary images, and segment all the targets, a new method of labeling connected components based on dynamic chained list was proposed. The line of target pixels was taken as the elementary cell. The pixel scanning and connected components labeling were processed synchronously. Simultaneously, the dynamic chained list were built. Based on the chained relation, only once pixel scanning was needed to segment all the targets. The fast open operation needed only once scanning to finish morphologic filter through a ＂hollow center＂ cross template and special judgment. Experimental results indicate that the dynamic chained list can solve the problem of labeling repeated effectively. For 512 ×512 images, The method needs only about 3 ms and is independent of the shape and number of the targets. Therefore, this method can meet the needs of high speed, exactness and robustness in engineering field.