白细胞图像的自动分割算法

Automatic segmentation algorithm for leukocyte images

根据白细胞核在其饱和度分量S和绿分量G中的分布特点,构造细胞变换图像,提取细胞核,实现白细胞的初步定位。对存在细胞粘连的细胞图像,利用细胞边缘的相位角变化来检测细胞粘连区域并除之,采用线性插值法拟合边缘曲线,得到封闭的白细胞边缘轮廓,最终提取白细胞。本算法可自动分割白细胞,具有很好的分割效果。

The paper constructs a transpositional cell image based on the saturation and green component images according to different distribution characteristic of leukocyte nucleus in the two images. The leukocyte nucleus can be extracted from the transpositional image and the initial location of the leukocyte can be achieved. For the overlapped cell images, firstly the overlapping regions are discovered from the phase change of the leukocyte contour and eliminated. Then the edge curve is re-constructed to get a closed contour using linear interpolation. Finally, the leukocyte can be extracted successfully. The proposed algorithm can be used to process the leukocyte automatically and achieve good results of leukocyte segmentation.