激光与光电子学进展, 2019, 56 (18): 181004, 网络出版: 2019-09-09   

基于Frangi滤波器和Otsu视网膜血管分割 下载: 1040次

Retinal Vessel Segmentation Based on Frangi Filter and Otsu Algorithm
作者单位
山东中医药大学理工学院, 山东 济南 250355
摘要
分析视网膜血管结构的变化是诊断和检测糖尿病、高血压等血管类相关疾病的最重要步骤。为此,提出了一种基于Frangi滤波器和大津法(Otsu)的视网膜血管分割方法。基于Frangi滤波器视网膜血管分割方法先对视网膜血管进行预处理,再基于Frangi滤波器对其边缘进行检测,并根据形态学方法分割出视网膜血管。然后,利用Otsu阈值分割预处理和增强后的视网膜血管。将上述两种方法获得的分割图像进行融合,获得最终的分割结果。实验结果表明,所提算法在DRIVE和STARE数据集上的平均灵敏性分别为58.1%、78.7%,特异性分别为93.1%、96.4%,而准确性则分别为93.8%、95.8%,其算法的执行时间仅为1.8 s。与传统的无监督分割算法相比,所提算法简单高效,能够很好地抑制噪声。
Abstract
Analysis of changes in retinal vascular structure is the key in the detection and diagnosis of vascular-related diseases, such as diabetes and hypertension. Herein, we propose retinal vessel segmentation methods based on the Frangi filter and Otsu algorithm. The Frangi filter-based method comprises the following steps. First, the preprocessing for retinal blood vessel is performed and then retinal blood vessel edges are detected using the Frangi filter; finally, retinal blood vessels are segmented using morphological approaches. The second method, based on the Otsu threshold segmentation technique, uses the retinal blood vessels after the Otsu threshold segmentation preprocessing and image enhancement. The final segmentation image is obtained by combining the outcomes of the above two methods. We apply the proposed algorithms on the two well-known DRIVE and STARE datasets. The average sensitivities obtained are 58.1% and 78.7%, the average specificities obtained are 93.1% and 96.4%, and the average accuracies obtained are 93.8% and 95.8%, respectively. The execution time of our method is 1.8 s. Results show that our algorithms are simpler and more efficient than the conventional unsupervised segmentation algorithm; in addition, they suppress noise very well.

孟琳, 刘静, 曹慧, 史婷瑶, 张驰. 基于Frangi滤波器和Otsu视网膜血管分割[J]. 激光与光电子学进展, 2019, 56(18): 181004. 孟琳, 刘静, 曹慧, 史婷瑶, 张驰. Retinal Vessel Segmentation Based on Frangi Filter and Otsu Algorithm[J]. Laser & Optoelectronics Progress, 2019, 56(18): 181004.

本文已被 7 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!