【摘要】 细胞图像分割是细胞图像识别过程中最困难也是最重要的步骤之一。本文针对临床使用的宫颈巴氏染色涂片细胞图像的特点,提出了二种新方法:自适应窗和去粘连法。在自适应窗内只有单个细胞存在,然后在窗内采用最大方差阀值法对该细胞进行分割;去粘连法避免了通常使用的腐蚀、膨胀次数难以确定的缺点,使粘连的细胞互相分离。经过病理学家对1100个细胞的测试,证明这二个方法对提高分割效果很有效。更多还原

【Abstract】 Two methods for segmenation on cell images:adaptive window and sepa-ration of touching cells are proposed in this paper.Corresponding algorithms are developed forpresercening Papanicolaou stamed cervical smear images obtained from black and whiteTV camera The adaptive window is different from the usual window with definite shapeand size.It is constructed adaptively according to the content of image to be analysed andthere is only one single cell in it without other dark artifacts which affect the result ofsegmentation.Then the thresholding method based on the maximal variance is appliedwithin the adaptive window to segment this single cell and the contours of nucleus and cytoplasm are extracted precisely.The method of separating touching cells is based on themorphological structure of cell,that is,the cytoplasm must surround the nucleus,and itsshape is similar to that of nucleus.This method overcomes the disadvantage of shrinking andexpanding operation for which the times of operation is difficult to be decided.The algorithmsare tested for segmenting 1100 cells by cytologists and good results are ob-tained.更多还原