目标分割图中粘连对象的自动切割和分离.pdfVIP

中国体视学与图像分析 2003 年3 月 第8 卷 第1 期 40 CHINESE JOURNAL OF STEREOLOGY AND IMAGE ANALYSIS Vol.8 No.1 Mar.2003 ·新技术与新方法· 文章编号：1007-1482(2003)01-0040-04 目标分割图中粘连对象的自动切割和分离 1 1 2 袁天云 ，姜志国 ，孟如松 （1.北京航空航天大学宇航学院图像处理中心，北京100083; 2.空军总医院皮肤科，北京，100036） 【摘要】针对图像分析中获得的目标分割图上出现的多个对象相互粘连的情况，根据此粘连对 象的形态特征，设计并实现了一种基于数学形态学水域生长的粘连对象自动切割和分离方法。该 方法通过在目标图中进行腐蚀和膨胀处理，由极限腐蚀获得各个独立对象的核，再由这些对象核 膨胀进行水域生长，产生分界线，最后达到自动切割、分离粘连对象的目的。本文给出了一种快 速、有效、自动地实现了粘连对象的分离处理方法，成功地解决了目标粘连现象对后续分析、测 量产生干扰的问题。 【关键词】粘连对象；对象核；水域生长；分界线 【中图分类号】TP 391.41 ，R445 【文献标识码】A AUTOMATIC SPLITTING AND SEPARATING OVERLAPPED OBJECTS IN TARGET-SEGMENTED IMAGE 1 1 2 YUAN Tian-yuan ， JIANG Zhi-guo ， MENG Ru-song (1.Image Processing Center, Beihang University, Beijing 100083 2.The Air Force General Hospital PLA, Beijing, 100036) 【ABSTRACTS 】In the target-segmented image, sometimes a few objects overlap together and this will interfere the subsequent analysis of the target. An automatic processing method based on mathematic morphology is presented in this paper to split and separate overlapped objects using watershed transform theory. The algorithm that involves erosion and dilation is designed for the purpose of automatic splitting and separating overlapped objects. It includes two steps: obtaining object-kernel according ultimate erosion and then giving dividing line by watershed expanding from the kernels. It is successful to solve the problem of overlapped objects with the high speed, the appropriate result and the automatic mode. 【KEYWORDS 】Overlapped object ；Object-ker