眼底图像硬性渗出物检测算法

Hard Exudates Detection Algorithm of Fundus Images

针对眼底图像中由于硬性渗出物堆叠造成的灰度分布曲线复杂,以及近距离血管割裂眼底背景造成误检等问题,提出了一种基于方向空间理论并结合上下采样平滑的眼底硬性渗出物检测算法。通过最大间距法,结合经过反向值修正的绝对幅值生成的直方图,对代表眼底硬性渗出物的凸线段进行提取。实验结果表明,本算法的敏感度、特异度和准确度在置信度大于50%区域炮较为理想。该算法符合医学诊断标准,可为相关疾病的诊断提供一种有效的辅助手段。

Aiming at the complex gray distribution curve caused by the stacking of hard exudates in fundus images and the false detection caused by the separation of fundus background by close blood vessels,an algorithm for detecting hard exudates in fundus based on direction space theory combined with up and down sampling smoothing is proposed.Through the maximum spacing method,combined with the histogram generated by the absolute amplitude corrected by the inverse value,the convex line segment representing the hard exudates of the fundus is extracted.The experimental results show that the sensiti-vity,specificity and accuracy of the algorithm are ideal in areas with confidence greater than 50%.The algorithm conforms to the medical diagnosis standard and can provide an effective auxiliary means for the diagnosis of related diseases.