基于改进GAC模型的血细胞显微图像分割

Segmentation on Micrograph of Blood Cells Based on Improved GAC Model

研究血液细胞图准确分割问题,针对血液细胞繁多、形态各异,造成图像准确分割困难。为获得更准确的分割结果,克服在血细胞显微图像分割中应用的典型分割算法的不足,对滤波方法和反差参数估计两方面对测地线活动轮廓模型进行了改进。分割时首先应用最小核同值区算子代替传统边缘检测算子,并结合形态学处理获得初始轮廓,根据改进GAC模型进行曲线演化获得精细细胞边界,最后,基于距离变换和测地膨胀获得粘连细胞分界线,完成细胞分割。仿真表明,典型改进算法对分割出的细胞区域更准确,细胞边缘连续平滑,而且具有良好的抗噪能力。

To shoot drawbacks of current typical algorithms used for segmentation on micrograph of blood cells,this paper proposes an improved Geodesic Active Contour Model（GAC） on aspects of filtering and the contrast parameter.In the process of segmentation,an initial contour,acquired by combination of non-differential SUSAN and morphologic operations,is provided to the improved GAC model which evolves to reach refined true cell edges,at last,with distance transform to search centers of overlapped cells,boundaries between overlapped cells are found by geodesic dilation.The simulation indicates that the proposed algorithm acquires more integral area of cells and continuous smooth edges,and favorable insistence to noise disturbance.