超分辨成像已成为活细胞结构和功能成像的关键工具, 荧光分子定位是超分辨成像过程中不可缺少的步骤。从超分辨成像角度研究各种荧光分子定位算法性能具有重要的意义。选择5种典型的荧光分子定位算法: 质心法、广义质心法、高斯拟合、解线性方程组和极大似然法, 以定位精度和定位时间来评价所选择算法的性能。结果表明, 1) 高斯拟合、极大似然法和广义质心法能高精度对荧光分子定位, 不受荧光分子所在子区域提取的影响; 2) 质心法和解线性方程组法能应用于图像在线分析, 但定位精度较低, 受子区域提取影响较大; 3) 当两个荧光分子位于一个衍射斑时, 采用这5种算法的定位精度都会急剧下降。

Super resolution imaging technique has become a key tool in imaging the structure and function of living cells. Localization of single fluorescent molecule is an integral part of super resolution imaging. From the point of view of super resolution imaging, a rigorous comparison of the performance of the different algorithms under experimental conditions is necessary. In this paper, five commonly used localization algorithms: centroid method, generalized centroid method, Gaussian fitting, fluoroBancroft method, and maximum likelihood method are investigated, and how well these algorithms work is clarified. The results show: firstly, Gaussian fitting, generalized centroid method and maximum likelihood method are high localization precision method and have been slightly influenced by extracting sub-window. Secondly, centroid method and fluoroBancroft method could be applied in on-time imaging, however, they have low localization precision and are influenced by extracting sub-window. Thirdly, for the five algorithms, the localization precision is abruptly decreased when there are two molecules in a diffraction spot.