基于马尔可夫随机场的眼底图像血管分割研究

眼底图像血管分割是医用图像分割中较为复杂的一种,在目前的研究中存在分割精度低、效率不高等问题。提出基于马尔可夫随机场的眼底图像血管分割算法,根据眼底图像的特点构建马尔可夫随机场模型,提取H通道作为特征场参数,利用最大后验准则完成标号场更新,最终实现对视网膜血管的分割。算法通过眼底图像数据库DRIVE进行测试,结果表明:该算法平均准确度为0.954 6,平均敏感度为0.899 9,平均特异度为0.957 1,具有很好的分割效果,且运行稳定,计算方便快捷,具有鲁棒性。

VESSEL SEGMENTATION OF FUNDUS IMAGE BASED ON MARKOV RANDOM FIELD

Vessel segmentation of fundus image is a relatively complex method in medical image segmentation. There are some problems in current research,such as low segmentation accuracy and low efficiency. This paper proposed vessel segmentation algorithm of fundus image based on the Markov random field. According to the characteristics of fundus image,a Markov random field model was constructed. H channel was extracted as the parameter of feature field,and the labeling field was updated by using the maximum posteriori criterion. Finally,the retinal blood vessels were segmented. The algorithm was tested by a fundus image database DRIVE. The results show that the average accuracy,sensitivity and specificity of the algorithm are 0.954 6,0.899 9 and 0.957 1 respectively. The algorithm has good segmentation effect,stable operation,convenient calculation and robustness.