通过对有缺陷的光纤端面进行研究，提出了基于灰度图像阈值分割和椭圆边界点选取及拟合的方法来求取缺陷光纤端面几何参数,并利用合格光纤的包层边缘、纤芯边缘为同心圆的特性来提取缺陷光纤边缘点，利用此特性能有效去除干扰点。与常见的两种方法(基于灰度阈值分割及特征数据点采集的方法、基于Canny算子边缘检测和椭圆拟合的方法)进行了对比分析，结果表明，该方法可以更精确地计算出缺陷光纤端面的几何参数。

On the bases of the study on defective fiber ends, this paper proposes an improved method for obtaining the geometric parameters, which is based on grayscale image threshold segmentation and elliptic boundary point selection and fitting, and extracting the edge point of the defective fiber by using the feature that the qualified fiber cladding and core edges have the same concentric circles. This feature can effectively remove the interference point. Compared with the two commonly used methods that are based on grayscale threshold segmentation and feature data point collection and on Canny operator edge detection and elliptic fitting respectively, this method figure out the geometric parameters of defective fiber ends more accurately.