摘要
目的:由于细胞图像十分复杂,传统的基于像素或者边界的图像分割方法难以精确的实现细胞分割。因此,需要设计一种可以实现细胞图像精确分割的方法。方法:结合大津分割算法和主动轮廓模型的优点,设计出一种基于单水平集函数的细胞分割算法,首先对细胞图像大津分割,其结果作为水平集函数的初始值,然后使用迭代法对水平集函数演化。采用MATLAB对显微镜下获取的细胞图像进行试验,将本文改进后的算法与常规的算法进行了对比。结果:与传统的水平集分割算法相比,本文方法对细胞图像分割结果更加准确,迭代次数减少一半左右,因此分割时间也减少了一半左右。结论:结合细胞图像的结构特点,利用大津分割结果作为主动轮廓模型的初始值,可有效解决主动轮廓模型因为初始值设置不当导致的分割缺陷问题,水平集函数能够跟踪拓扑结构变化,具有计算精度高、算法稳定、优化边界清晰光滑等优点,在本文中得到了充分的应用。因此本文所提出的算法能够高效地实现细胞图像的分割。
Objective： To design a method this could achieve accurate cell image segmentation. Methods： Because of the com- plexity of cell image, traditional image segmentation methods based on pixel or boundary could not get good results. In this pa- per, taking the advantages of the Otsu segmentation algorithm and active contour model, a cell segmentation algorithm based on single level set function was proposed. First, the cell image was segmented using Otsu method, the initial value of level set func- tion was set based on Otsu segmentation results, and then the level set function was iterated. The cell images got from micro- scope were processed using MATLAB. The improved algorithm and the conventional algorithm were compared. Results： Com- pared with traditional level set segmentation algorithm, this method was more accurate for image segmentation. The iterations were reduced by about half and so it was the processing time. Conclusions： Considering the structure characteristics of cell im- age, the segmentation problem of active contour model because of inappropriate initial value were resolved efficiently using the results of Otsu as initial value. Level set fimction can track topology changes and has the advantages of high accuracy, stable al gorithm and boundary optimization. This proposed algorithm could achieve cell image segmentation more efficiently.
出处
《中国医学物理学杂志》
CSCD
2013年第6期4523-4526,4592,共5页
Chinese Journal of Medical Physics
基金
山东省自然科学基金面上项目(ZR2012HM060)
国家自然科学基金(81171330)
中央高校基本科研业务费专项资金(12CX04076A)
