一种新的视网膜血管图像增强方法  

A novel method for retinal vascular image enhancement

作　　者：廖苗[1] 郑少伟[1] 赵于前[1] 

机构地区：[1]中南大学地球科学与信息物理学院,湖南长沙410083

出　　处：《光电子．激光》2012年第11期2237-2242,共6页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61172184);湖南省自然科学基金(12JJ6062);中南大学中央高校基本科研业务费专项基金(2011QNZT013)资助项目

摘　　要：提出了一种基于多尺度顶-帽变换和直方图拟合拉伸的视网膜血管图像增强方法。首先,应用多尺度顶-帽变换提取不同尺度下的图像亮、暗特征,并从中选择最优的特征对图像进行初步增强。然后,结合视网膜血管图像的灰度直方图分布特性,对初步增强结果进行直方图高斯曲线拟合线性拉伸。为验证方法的有效性,以Drive眼底图像数据库为实验对象,并与其它多种增强算法进行了比较。实验结果表明,相比其它算法,本文方法能够更加有效地提高图像的对比度,并在细节上有很好的增强效果。Retinal vessels play an important role in the diagnostic procedure of retinopathy. A new retinal vascular image enhancement method is proposed. Firstly,the retinal image is enhanced preliminarily by using the optimal image features which are selected from the multi-scale white and dim image features extracted by multi-scale top-hat transformation. And then according to the histogram distribution charac- teristics of retinal vascular gray image, the preliminary enhanced image is processed by linear stretching with histogram Gaussian curve fitting. The experimental results show that the proposed method im- proves the contrast and enhances the details of the retinal vascular images more effectively than the ex- isting methods.

关 键 词：图像增强 数学形态学 多尺度顶-帽变换 线性拉伸 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]