基于测地形态学进行细胞图象分割的新方法

New Method of Segmentation on Cell Image Using Geodesic Morphology

细胞分割提取在计算机病理诊断系统中是一个重要技术,分割结果的优劣直接影响着诊断的效果,也是影响诊断系统能否应用在临床上的一个重要因素。本文使用数学形态学的测地形态学方法,利用细胞图象的灰度图特征,提出一种有效的细胞图象分割方法,对原始图象进行中值滤波预处理后,首先使用水浸分割算法得到单个细胞的分割结果和聚堆细胞的外轮廓,然后使用水线分割算法得到水线图,最后通过寻找细胞外轮廓的凹点,结合水线图,实现了聚堆细胞的分离。实验结果表明方法准确可靠,具有一定的通用性。

Cell segmentation is an important techinque in computer diagnose system.A new method based on geodesic morphology is proposed,utilizing the grayscale characteristic of cell image.First,the median filter is used on the input image ,then the segmentation results of separated cells and the outer edge of overlapping cells are obtained by using water immersion segmentation algorithm.Next the watershed image is obtained by using watershed segmentation algorithm.And the cancave dots on the outer edge of overlapping cells are found.Finally the overlapped cells are separated by combining the cancave dot and watershed image.The result of segmentation is exact and reliable,and tests show the method has strong applicability.