摘要
夜视系统已经成为驾驶辅助系统的重要组成部分,但夜间采集的图像在大面积低照度和对面来车强光干扰下表现出明显的光照不均匀性。提出了一种基于多尺度tophat变换和虚拟融合的图像增强和眩光抑制方法,首先采用多尺度tophat估计光照曲面,获得目标图像;进而将目标图像叠加不同的虚拟光照获得多幅光照补偿的多幅图像,根据对比度和曝光品质的指标进行多分辨率融合。实验表明该方法实现了眩光抑制和暗区增强,有效解决了夜视图像光照不均匀性问题,使得处理后图像最远可以辨识200 m处的目标。
Night vision system is become more useful in advanced driver assistance system.Night vision images have the imbalanced illumination at large area low light and the glare from opposite car.This paper presents image enhancement and glare suppression method based on multiscale top hat transform and virtual fusion.Firstly,the method estimate illumination surfaces and get the target image using multiscale top hat transform.Then,it makes different compensation images through superposition between the target image and the different virtual illumination.Multi resolution images fusion is made by the contrast and exposure quality measures.Experiments show that the dark area of image is enhanced and the glare is suppressed.The problem of imbalanced illumination is solved effectively in night vision.The enhancement image can recognize farthest the target at 200 meters.
出处
《电子测量与仪器学报》
CSCD
2014年第4期368-372,共5页
Journal of Electronic Measurement and Instrumentation
基金
国家博士后基金面上项目资助(2013M531504)
关键词
图像增强
眩光抑制
tophat变换
虚拟融合
image enhancement
glare suppression
tophat transformation
virtual fusion
