一种基于显微多光谱宫颈细胞图像自动分割方法

介绍了一种基于多光谱宫颈显微图像自动分割方法。细胞涂片每一个局部区域经过显微镜放大后再通过电调谐光器件分别取出不同光谱波段的图像，经过CCD送入计算机后得到该视场分析所需的多光谱图像。从所获得的多光谱图像中选择一定的波段图像对数变换相除、多阈值分割和形态学操作，最后获得宫颈细胞的胞浆和胞核覆盖层。本文首次将多光谱分析应用到宫颈显微图像分割中，具有分割准确、分割速度快、能够去除胞浆重叠和受外界干扰少的特点。

Automatic Segmentation Method for Cervical Cell Based on Multispectral Microgram

An automatic segmentation method for the cervical cell based on multispectral microgram is presented. The segmentation procedure consists of three phases. Firstly, a novel method is proposed for obtaining multispectral images of cervical cells. In the second phase, two new grey images are generated, one for cytoplasm and the other for nuclear. In this phase, some certain bandwidth images are selected, and a logarithmic transformation is made for these selected images, then the transformation images are divided to generate new images. Finally, new grey images are separated with aumatic thresholding, and the segmented regions are smoothed by morphological operations. Experimental results show that the approach for segmenting cervical images is rather accurate and fast and can eradicate the influence of overlapped cells. It is robust to the noise.