一种免疫细胞图像非监督分割方法.PDFVIP

中国体视学与图像分析 2002 年12 月 第7 卷 第4 期 224 CHINESE JOURNAL OF STEREOLOGY AND IMAGE ANALYSIS Vol.7 No.4 Dec. ·新技术与新方法· 文章编号：1007-1482 （2002）04-0224-05 一种免疫细胞图像非监督分割方法 谢凤英，姜志国 （北京航空航天大学宇航学院图像中心，北京100083） 【摘要】对免疫细胞图像进行全自动分析在医学疾病诊断上有很高的应用价值。要完成自动图 像分析，关键是要实现图像的非监督分割。根据免疫细胞图像中细胞核是一种椭圆状目标的特点， 提出了一种椭圆形目标定位方法。在文献中，首先用这种目标定位方法将待分割图像中的细胞核 位置找到，然后在细胞核所在的区域内运行水域分割算法（watershed变换）就可实现图像的非 监督分割。这种非监督分割策略解决了在大范围内运行中算法一次分割目标时存在的容易错分割 的缺点，实现了对特定细胞图像的非监督分割。实验结果表明该方法具有目标定位准确率较高、 分割的准确精度和可重复性好等特点。 【关键词】水域分割；非监督分割 【中图分类号】Q271 【文献标识码】A AN UNSUPERVISED SEGMENTATION METHOD FOR IMMUNE CELL IMAGE XIE Feng-ying, JIANG Zhi-guo (Image Processing Center. Beijing University of Aeronautics and Astronautics. Beijing 100083) 【ABSTRACT 】Automatically analyzing of immune cell image can be used for diagnosing the illness and has highly value in medical science. And automatically segmenting image is the crucial step for this analysis . In this paper, an ellipse object location detecting method is described. With this method the location of ellipse karyon can be found in the immune cell image. After having gotten the location of each karyon, watershed transform is run in each little region which karyons exists in. As a result, the image was automatically segmented. The immune cell image would often be segmented falsely if the segmentation method described in [1] is run in a large region. By using this unsupervised segmentation method the problem can be solved very well. Through experiment the locating method for ellipse object is proved to be highly precise, and the unsupervised segmentation result can meet the need. 【KEYWORDS 】watershed segmentation;