基于结构特征的视网膜血管形态识别

Pattern recognition of retinal blood vessels based on nodes structure feature

针对视网膜眼底图像获取过程中眼球转动的问题,提出一种基于节点最近邻结构的具有旋转和平移不变性的视网膜血管形态识别方法。该方法利用节点的周边结构稳定性的特点来进行节点结构特征提取,进行图像相关结构匹配的判定。实验结果表明了该识别算法的有效性和可靠性,且不需要进行方位对准处理,具有较好地识别灵活性和实用性,正确识别率达到98.57%。

In order to overcome the effects of eye movement during the funds image acquisition stage, a method based on the node＇ s adjacent structure for pattern recognition of retinal blood vessels is proposed. The method has the quality of rotation and translation invariance. Firstly, the node structure characteristics are extracted using the structural stability of the surrounding nodes, and then matching the image correlation structure. Experimental results on a database demonstrate the effectiveness and reliability of the recognition algorithm. It does not need to the complex pre-proeessing task for level alignment, and has a better adaptability and practicability. The accuracy rate of 98.57% is reached for the proposed method.