Abstract: Cell images processing in medical often appear that the cell images have overlapped each other.The traditional watershed algorithm always results in serious over segmentation phenomenon.It can avoid over segmentation well by using the distance transformation watershed algorithm.Considering that cells have a similar circular characteristic in common，doing the Ultra-erosion by using the fixed structuring element can result in cells’ distortion easily.Eroding the overlapping cells by using the alternant structuring element can mark every single cell seed accurately，and it can avoid the cells’ distortion by eroding effectively.Finally，it finds out the overlapping cells’ watershed by using the dilation.The results show that the algorithm can avoid over segmentation effectively and improve the measurement accuracy.

Key words: overlapping cells, watershed algorithm, distance transformation, ultra-erosion

摘要： 在医学细胞图像处理中，经常遇到细胞图像重叠现象。传统的分水岭算法存在严重的过分割现象。使用距离变换的分水岭算法能很好地避免过分割现象。考虑到细胞具有类圆形的特点，使用固定的结构元素进行极限腐蚀容易导致细胞失真。使用交替结构元素对重叠细胞进行腐蚀，能准确地标记出单个细胞种子，同时能避免细胞因腐蚀而失真。最后使用膨胀处理找出重叠细胞的分界线。实验结果表明，算法有效地避免了过分割现象，并提高了分割处理的准确度。

关键词: 重叠细胞, 分水岭算法, 距离变换, 极限腐蚀