中文摘要：

提出了一种分割老年黄斑变性视网膜频域光学相干断层图像的方法．该方法通过如下技术在保持高精度的同时减少计算时间：首先采用费舍尔判别分析方法对视网膜各层分界面进行粗分割，其次采用曲率计算方法对玻璃膜疣进行检测，最后使用卡尔曼滤波优化分割效果．该方法对20卷体数据中220幅老年黄斑变性图像的三层分界面进行分割验证，在平均绝对误差小于3．29μm的同时，每幅平均处理时间小于42ms．与代表当前最好水平的文献相比，本文所提出的算法能在保持精度的基础上将处理时间缩短40倍，因而能更好地适应于临床需求．

英文摘要：

An approach to segment ocular optical coherence tomography images with age-related macular degeneration was developed. The main advantage is to decrease computational loads while maintaining high accuracy through the following techniques： Fisher＇s discriminant analysis for initial location of layer interfaces, curvature calculation for drusen detection, and refining interfaces using Kalman filters. Validation on 220 images of 20 volumes shows that three layer interfaces in each image can be segmented within 42 milliseconds with an average absolute error of layer interface below 3.29 μm. Compared with a state-of-the-art method, the proposed method is 40 times faster and maintains similar or significantly better accuracy,which is better suited for clinical usage.