彩色骨髓相的快速分割

为了利用计算机进行髓相分析,对在彩色空间中复杂背景下的目标分割问题作了研究。提出了在处理后的色度和饱和度分量的融合图像上,根据改进的最大熵二值化算法在两个层面上对细胞核进行多级分割,解决复杂背景下细胞核及粘连细胞核的分割问题。再由细胞核外周对空间作搜索,根据统计方法确定的点集性质一次性地对细胞胞浆做分割,最后分析细胞的几何特征,对分割进行调整。实验结果表明:整个算法适应性强,对不同性质的骨髓图像均有较好的分割效果,且与其他血细胞分割方法相比计算量小,比较快捷,约86%的细胞得到了正确分割。

Rapid segmentation of chromatic bone marrow cell images

Many systems have been developed to computerize the analysis of bone marrow cells. The paper presents a method to segment objects in a complex chromatic background. The algorithm uses a level-set segmentation of the nucleus and other parts in the image based on an enhanced hue and saturation sub-image calculated using an improved maximum entropy algorithm. The algorithm then searches the space outside the nucleus to statistically select plasma pixels and then segment the plasma based on the pixel characteristics. The algorithm works well with various kinds of bone marrow images and is quicker than other blood cell segmentation methods. The segmentation accuracy is about 86%.