荧光视网膜图像的照度均衡及自适应血管增强算法  

Illumination equalization and adaptive vessel enhancement in fluoroscein retinal images

作　　者：刘尚平[1,2] 陈骥[1] 

机构地区：[1]重庆大学生物工程学院,重庆400044 [2]香港中文大学影像与介入放射学系

出　　处：《光电子．激光》2011年第5期793-797,共5页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(30970764);重庆市科技重大专项资助项目(CSTC2008AA5010)

摘　　要：提出一种基于图像反射度照度模型的荧光视网膜图像照度校正与均衡算法。首先,使用多尺度处理及形态学测地膨胀运算将原始图像分离为背景图像和前景图像,然后,在背景图像的基础上使用多方向直线均值法对图像的照度成分进行估计,并由反射度照度模型得到背景的均衡图像;最后,单独对前景血管自适应增强的图像进行灰度校正,并将均衡后的背景图像与校正后的前景图像结合起来得到最终的均衡图像。对30幅眼底荧光造影视网膜图像进行测试,结果表明,本文算法不仅消除了图像中的照度不均现象,有效地保留了图像内部的主要特征,而且在处理效果及性能上均优于同态滤波、低通校正(LP)等主流的照度校正和均衡方法。An efficient illumination equalization approach based on illumination and reflectance model is proposed.The method started by separating the original image into the background image and foreground image using multi-scale processing and geodesic dilation.Then the illumination of the background image was estimated by multi-directional line-average method and the equalized background image was computed by dividing the image by illumination component.Finally,gray scale correction was applied for the foreground image,and the equalized background image and the corrected foreground image were integrated to realize global equalization.The proposed method was evaluated using database with 30 images.The experiments demonstrate that this method can estimate the detrimental effect of non-uniform illumination while persevering main features well,and the performance and visual effect of the proposed method are better than those of many mainstream illumination correction methods,such as homographic filtering and low-pass correction.

关 键 词：视网膜图像 照度 均衡 血管 增强 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]