针对人眼无意识的抖动导致图像位置在眼底成像设备上无规律变化而无法有效观察指定区域的问题，提出了一种视网膜图像实时动态稳像算法.该算法依据图像匹配理论，在图像探测的同时引入图像有效区域实时“计算移位”.利用时间上相邻的视频序列图像近似对应于眼底同一位置的特点，通过“采集运算”找到当前采集图像的中心区域.把当前图像的中心区域与第一帧采集图像的中心区域进行相关匹配，获取当前视网膜图像相对第一帧图像抖动的偏移方向和偏移量，然后把当前图像移位，循环执行此过程，并剔除无法匹配的图像，使得屏幕上的图像始终是在同一个位置的静态图像.实验证明：本文提出的算法解决了自适应眼底成像设备在对视网膜细胞观测时因为人眼自然抖动而造成的图像不稳定问题;解决了人眼在可见光波段的激光照明下受到较强刺激而无法长时观测的问题，实现了细胞和微血管同平台同算法观测;是采用拼接技术实现大视场图像的前提.

The unconsciousshake of human eye leads to the random variations of image position, and makes it impossible to observe a specific area. To solve the problem, during the object imaging, a real time “calculationmovement”stabilizing image algorithm was presented based on image matching theory. It uses a series of continuously grabbed images that focus on nearly the same area of retina. Therefore, it is possible to find the center area of the image through “samplingcalculaiton”, and do the correlation between the center area and that of the first image so as to get the direction and amplitude of movement, finally move the picture according the results. Repeat this process to stabilize the series of images and remove those unmatched images, which make the object image nearly on the same position. The results indicates that the algorithm can overcome the random variation problem of object image position and stabilize the retinal image in real time for the adaptive optics imaging setup; overcome the problem that one cannot focus on the same area for a long time because of laser′s strongly irritation, and realize the retina and blood capillary observation with the same setup and same algorithm; make it possible to get anpatched image with large field of view.