【摘要】 细胞图象的分割目的在于按亮度(灰度)从图片的背景中提取出单个细胞图象,并划分出背景、胞浆和胞核三个区域以便对有关区域抽取特征。分割的质量将对特征的抽取与最终的判别分类结果起直接的影响。 本文试验所采用的是一组包括有癌、重增、轻增和正常四种类型的88个食管上皮细胞样品。 本文的目的是研究如何将阂值聚类、边界检测、区域生长三种技术具体应用于细胞图象的自动分割。更多还原

【Abstract】 <ABSTRACT>The purpose of segmentation is to separate a single cell from background and to segment those three regions belonging to background, cytoplasm and nucleus according to the grey level values of each pixel, so that the featu res can be extracted for those regions-In this paper, the automatic segmentation of cell images by meansof thresholding technique, edge detection technique, and region growing tech nique are described and tested on a set of 88 esophagus epithelial cells-The experimental results indicate that thresholding method exhibits quite simple and efficient for most of the common cell images; edge dete ction method may be more efficient for some individual cell images; and region growing method can be used in cell image segmentation, but its alg orithm usually involve great expenditure in computation time and memory requirments-In order to get a more reasonable solution for cell image segm entation, one of the research directions will be to combine the edge detect-tion -with thresholding technique.更多还原