改进的形态学与Otsu相结合的视网膜血管分割

针对视网膜图像采集过程中由于疾病引起的图像光照反射过强问题,提出了一种修正的形态学与Otsu相结合的无监督视网膜血管分割算法。首先运用形态学中的高低帽变换增强血管与背景的对比度;然后提出了一种修正方法,消除部分由视网膜疾病引起的光照问题;最后使用Otsu阈值方法分割血管。算法在DRIVE和STARE视网膜图像数据库中进行了测试,实验结果表明,DRIVE数据库中的分割精度为0.9382,STARE数据库中的分割精度为0.9460,算法的执行时间为1.6s。算法能够精确地分割出视网膜血管,与传统的无监督视网膜血管分割算法相比,算法的分割精度高、抗干扰能力强。

New approach to segment retinal vessel using morphology and Otsu

The appearance and structure of retinal vessels play an important role in the diagnoses in ophthalmic diseases, it also plays an important role in the diagnosis of cardiovascular disease and diabetes, which requires an algorithm to automatically extract the retinal vessels. To enhance the image with imbalanced local illumination caused by retinal diseases in the process of fundus image acquisition, this paper proposed a new unsupervised retinal vessel segmentation method with morphological and Otsu. First, used the combine of the top-hat transformation and the bottom-hat transformation to enhance the contrast between the blood vessels and its background in a retinal image. Next, presented a novel revised method to remove the problem, which is caused by the imbalanced local illumination in the enhanced retinal image. Finally, applied the threshold calculated by Otsu method to extract the retinal vessels. We evaluated the algorithm with two publicly retinal image databases DRIVE and STARE. The experiment results indicate that the segmentation accuracy in DRIVE database achieves 0.9382, the segmentation accuracy in the STARE database achieves 0.9460. The run time of our new method is 1.6 seconds. The new algorithm can accurately extract the retinal vessels. Compared with the traditional retinal vessel segmentation algorithm, its segmentation accuracy and anti-perturbation ability improve.