The extraction of blood vessels helps in the diagnosis of diseases and to develop advances of medicine. Retinal blood vessel extraction plays a crucial role in early detection and treatment of retinal diseases. This paper provides an automatic segmentation of blood vessels in retinal images. First, the fundus images go through preprocessing steps of image acquisition, grey scale conversion, bias correction and adaptive histogram equalization to enhance the appearance of retinal blood vessels. Then the retinal blood vessels are extracted using a probabilistic modeling and maximum entropy based expectation maximization algorithm which uses maximum entropy uniform distribution as the initial condition. The vessels are more accurately confined using image profiles computed perpendicularly across each of the detected vessel centerline. The algorithm is implemented in MATLAB and the performance is tested on retinal images from DRIVE and STARE databases. When validated, we conclude that the segmentation of retinal images using the proposed method shows a sensitivity of 98.9%, a specificity of 83.74%, and an Accuracy score of 98.8%.

提取血管有助于疾病的诊断和医学发展。视网膜血管提取在视网膜疾病的早期发现和治疗中起着至关重要的作用。本文提供了视网膜图像中血管的自动分割。首先，眼底图像经过图像采集，灰度转换，偏差校正和自适应直方图均衡化的预处理步骤，以增强视网膜血管的外观。然后使用概率模型和基于最大熵的期望最大化算法提取视网膜血管，该算法使用最大熵均匀分布作为初始条件。使用跨每个检测到的血管中心线垂直计算的图像轮廓可以更精确地限制血管。该算法在MATLAB中实现，并且对来自DRIVE和STARE数据库的视网膜图像进行了性能测试。经验证后，我们得出的结论是，使用所提出的方法对视网膜图像进行分割显示出98.9％的灵敏度，83.74％的特异性和98.8％的准确性得分。