摘要
在追踪序列图像中大量疏密不均的细胞时,由于细胞存在粘连、分裂、移入与移出等问题,影响其分割与追踪的准确性。为了进一步提高追踪的准确性,提出一种蜂窝划分的多帧及原图反馈修正的细胞追踪方法。首先把分割好的图像进行蜂窝状区域划分,然后引入Delaunay三角网建立细胞邻域图,再应用拓扑约束实现细胞匹配,最后把细胞共分为欠分割、过分割、粘连、分裂、移入、移出和未处理细胞共七大类,并提出多帧反馈和原图反馈法对匹配结果和分割错误进行修正。实验结果表明,本方法能够有效追踪细胞分布疏密不均、团簇问题严重及细胞于帧间游动速度快的图像序列,不但大幅度地修正分割错误以降低追踪对分割的依赖性,而且可拓宽拓扑约束法的适用性,最终使追踪准确率有很大的提高。算法测试了3个细胞图像序列,比拓扑约束法的追踪准确率分别提高25.61%、77.14%和45.83%。
When tracking a large number of cells in image sequences which distribute in uneven density,due to cells’ cluster,division,moving in,moving out and so on,the accuracy of segmentation and tracking is decreased significantly.To improve the accuracy,a novel cell tracking method was presented in this work,which was based on cellular partition combined with feedback correction of multi-frames and original images.We firstly partitioned all segmented images into cellular region,then applied Delaunay triangulation to establish cells neighborhood graph.After all cells were matched via topology constraints,we classified all cells into seven categories which were under segmentation,over segmentation,cluster,division,moving in,moving out and obscure one.We applied feedback of multi frames and original images to correct matching and segmentation errors finally.Experimental results showed that the method effectively tracked the image sequences with cells’ uneven density distribution,seriously clustering and moving fast.Three image sequences were tested.Compared with the topology constraints algorithm,the tracking accuracy was imcreased 25.61%,77.14% and 45.83% respectively.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2012年第3期396-405,共10页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金(60875020)
关键词
细胞追踪
拓扑约束
蜂窝区域划分
DELAUNAY三角网
多帧及原图反馈修正
cell tracking
topological constraints
cellular partition
delaunay triangulation
multi-frames and original image feedback correction
