病变视网膜图像的血管骨架提取方法研究

Extracting blood centerline adapted for retinal fundus images with pathologies

现有血管提取方法很少考虑到视网膜病变后,受病变区域边界干扰,很难正确提取血管结构。针对眼底视网膜病变图像,利用不同相位下Gabor滤波函数对不同特征模式的响应特性,提出一种基于Gabor滤波的非血管结构抑制血管提取方法,同时结合非极大值抑制细化以及滞后多阈值处理技术,最终实现眼底图像血管骨架提取。实验结果表明,该方法优于同类高斯匹配滤波方法和不采用非血管结构抑制的Gabor滤波方法。

Existing blood vessel extraction methods hardly regard to retinopathy image.Due to the interference of lesions border,it is difficult to correctly extract the vasculature.In this paper,we propose a novel method based on differ-ent phase offsets which correspond to different responses of Gabor functions against image feature.The method applies non-vessel structure inhibition based on Gabor filter,and combines with non-maxima suppression and multi-threshold hysteresis technique.Experimental results show that our method is better than Gaussian matched filter and Gabor filter without non-vessel structure inhibition.