基于形态学的均值平移算法的神经元干细胞图像的自动分割

在神经元干细胞的图像分析中,准确快速的图像分割是干细胞分化增值自动追踪系统的基础。为了准确地分割低对比度的灰度神经元干细胞图像,本研究提出一种基于形态学运算和均值平移算法的神经元干细胞分割方法,称其为形态学的均值平移算法。此算法可以快速地获得任意形状细胞的图像,并且能检测到图像中多连接边缘不封闭的细胞。将此方法应用于神经元干细胞序列图像分割中并且将其与门限分割、水线分割和活动轮廓进行对比。实验结果证明,与其它的方法相比,此方法获得的细胞分割形状更接近于真实细胞的形状,并且能获得或接近于原始图像中准确的独立细胞数目。此方法可以获得正确的分割结果,为进一步图像处理奠定了良好的基础。

Aultomatic Segmentation of Neural Stem Cell Image Based on Morphologocal Mean Shift

In the analysis of neural stem cells, accurate and rapid segmentation is the base of automatic tracking system. To segment the low contrast gray-level image of adult neural stem cells, this paper presented a new segmentation method, integrating mathematical morphological operation with mean shift method. This method, so called morphological mean shift method, can rapidly obtain the arbitrary shapes of cells and divide complex cells into single ones. The new method was applied in sequential images of neural stem cells and compared with other partition methods like threshold, watershed and active contour. Comparisons proved that the proposed method could obtain better cell shape and more precise number of cells. The proposed method can get the accurate segmentation results, laying a solid foundation for further image processing.