中文说明：本文提出了基于特征的视网膜图像配准的一种新的结构特征，叉结构是由一个主分叉点和它的三个连接的邻居构成。每个分叉结构的特征向量组成的归一化分枝角度和长度，这对平移，旋转，缩放，甚至适度的失真是不变的。只要血管图案可分段这样可以大大减少了匹配过程的不适定性。该方法的简单性和效率，使得它很容易被单独使用或与其他现有方法纳入到制定混合或分层方案。
English Description：
Retinal image registration based on feature is presented in this paper, a new structure of fork structure consists of a main point of bifurcation and its three connecting neighbors make. Each bifurcation structure of normalization of characteristic vectors branching angles and lengths, the translation, rotation, scaling, and even a moderate amount of distortion is the same. As long as the vascular pattern can be segmented so that can greatly reduce the uncertainty of the matching process. The simplicity and efficiency of this method makes it very easy to be used alone or in combination with other existing approaches to designing mixing or layering scheme.