基于相位一致性的眼底图像血管分割算法

Retinal image vessels segmentation algorithm based on phase congruency

眼底图像的血管分割对糖尿病的分析和诊断至关重要.为了准确地分割出眼底图像中的血管,本文提出了一种基于相位一致性的眼底图像血管分割算法.本算法首先采用对比度受限的自适应直方图均衡化和各向异性耦合扩散方程对图像进行预处理,然后利用相位一致性算法来提取眼底图像中的血管,最后用数学形态学对其进行优化.利用本文提出的算法对国际上公开的Hoover眼底图像库进行测试,本算法的准确度可达到94.36%,实验结果证明了本算法的有效性.

Retinal image vessels segmentation is very important for analysis and diagnosis of diabetes mellitus（DM）.To achieve vessel segmentation with high accuracy,a retinal image vessels segmentation algorithm based on phase congruency is introduced.The contrast-limited adaptive histogram equalization（CLAHE） and the anisotropic coupled diffusion equations are used for image preprocessing.Then the vessels are extracted using phase congruency（PC）.At last,morphological operation is applied for refinement.This algorithm has been tested on the public Hoover database and the accuracy ratio is 94.36%.The experimental results prove the high performance of the proposed algorithm.