基于过渡区提取的视网膜血管分割方法  

Segmentation of Retinal Blood Vessels Based on Transition Region Extraction

作　　者：姚畅[1] 陈后金[1] 李居朋[1] 

机构地区：[1]北京交通大学电子信息工程学院,北京100044

出　　处：《电子学报》2008年第5期974-978,共5页Acta Electronica Sinica

基　　金：新世纪优秀人才支持计划(No.50051);教育部科技重点项目(No.106030)

摘　　要：针对现有视网膜血管分割方法对于小血管和低对比度血管分割效果差的问题,提出了一种基于过渡区提取的视网膜血管分割方法.该方法首先采用二维高斯匹配滤波预处理以增强血管,然后采用基于最佳熵的方法提取主血管、采用基于分布式遗传算法和Otsu相结合的方法提取过渡区,最后利用区域连通性分析所提取的主血管和过渡区,分割出最终的血管.通过在Hoover眼底图像库中的实验,结果表明该方法在小血管的提取、连通性和有效性方面均优于Hoover算法,另外由于迁移策略的分布式遗传算法的引入,使得算法效率也明显提高.Aiming at the bad performance of existing retinal blood vessel segmentation methods for small and low contrast vessels, a new segmentation method based on transition region extraction is proposed. Firstly, the two-dimensional Gaussian matched filter is used to enhance the retinal image. Then the main vessels are segmented through optimal entropy method and the transition region is extracted by algorithm based on distributed genetic algorithm and Otsu. Finally, the vessels are obtained via analyzing the region connectivity. The experiments implemented on the Hoover database indicate that the proposed method outperforms the Hoover algorithm on the small vessels extraction,connectivity and effectiveness. In addition,the efficiency of this method could be improved by introducing the distributed genetic algorithm based on migration strategy.

关 键 词：血管分割 过渡区 最佳熵 分布式遗传算法 最大类间方差 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]