首页|基于数字衍射的单幅眼底图像增强

基于数字衍射的单幅眼底图像增强

扫码查看
彩色眼底图像是分析与监控眼底疾病的重要工具。由于照明不均匀的问题,眼底图像视觉质量不足,图像对比度有待提高。为此,基于数字衍射提出了一种兼顾颜色保真与亮度增强的单幅眼底图像光强校正算法。将彩色RGB眼底图像转换到LCH色彩空间进行基于L通道的光强校正,以解决眼底图像亮度的平衡问题。对C通道进行相同操作使得处理后的彩色眼底图像颜色保真性较好。在Messidor眼底图像数据集的1 200组眼底图像上进行分析,并与Gamma校正、Retinex等眼底图像光强校正方法进行对比。本算法不仅具有更好的图像增强效果,改善了彩色眼底图像的颜色失真及对比度低的问题,还可进一步结合自适应直方图均衡化(CLAHE)以提升图像对比度。实验结果表明,在图像多尺度对比度和图像噪声评价指标方面优于传统算法3%~4%。经算法增强后眼底图像亮度分布更加均匀、对比度提升,为后期眼底图像病理位点的识别、血管与病灶分割和分类提供了一种性能更好的预处理方法。
Digital diffraction method for single retinal image enhancement
Enhancing RGB retinal images is vital for retinopathy detection and monitoring,but issues like uneven intensity often degrade visual quality.This research introduces a digital diffraction-based method to improve uneven intensity and contrast while preserving color naturalness.Initially,the retinal image is con-verted to LCH color space,where intensity correction is applied to the L channel,treated as an optical field.A digital propagation with a specific kernel estimates the intensity pattern,which,when subtracted,yields a corrected L channel.Multi-image fusion with varied kernels then ensures uniform intensity.The same process corrects the C channel for color accuracy.Tested on 1 200 Messidor dataset images,this method surpasses Gamma correction and Retinex methods,enhancing contrast and uniformity by 3%-4%when combined with CLAHE.The improved contrast aids applications like retinopathy detection and blood vessel segmentation.

medical image processingretinal image enhancementcontrast enhancementdigital diffrac-tion

张书赫、曹良才

展开 >

清华大学 精密仪器系,北京 10084

医学图像处理 眼底图像增强 对比度增强 数字衍射

国家重点研发计划资助项目

2021YFB2802300

2024

光学精密工程
中国科学院长春光学精密机械与物理研究所 中国仪器仪表学会

光学精密工程

CSTPCD北大核心
影响因子:2.059
ISSN:1004-924X
年,卷(期):2024.32(15)