[1]孙淑杰,董燕成,汤春明,等.单幅夜间低照度车载视频图像的增强算法[J].计算机技术与发展,2018,28(04):50-54.[doi:10.3969/ j. issn.1673-629X.2018.04.011]
　SUN Shu-jie,DONG Yan-cheng,TANG Chun-ming,et al.An Enhanced Algorithm for Single Nighttime Low Illuminated Vehicle-mounted Video Image[J].,2018,28(04):50-54.[doi:10.3969/ j. issn.1673-629X.2018.04.011]
点击复制

单幅夜间低照度车载视频图像的增强算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
28
期数:
2018年04期
页码:
50-54
栏目:
智能、算法、系统工程
出版日期:
2018-04-10

文章信息/Info

Title:
An Enhanced Algorithm for Single Nighttime Low Illuminated Vehicle-mounted Video Image
文章编号:
1673-629X(2018)04-0050-05
作者:
孙淑杰;  董燕成;  汤春明;  商大伟;  匡澄;  常继英
天津工业大学
Author(s):
SUN Shu-jie; DONG Yan-cheng; TANG Chun-ming; SHANG Da-wei; KUANG Cheng; CHANG Ji-ying
School of Electronics and Information Engineering,Tianjin Polytechnic University,Tianjin 300387,Chin
关键词:
夜间低照度;  图像增强;  优化;  单幅车载视频图像
Keywords:
nighttime low-illumination; image enhancement; optimization; single vehicle-mounted video image
分类号:
TP391.1
DOI:
10.3969/ j. issn.1673-629X.2018.04.011
文献标志码:
A
摘要:
针对目前车载HUD(head up display)系统未对低照度路况做出相应处理的问题,提出一种适用于单幅夜间低照度车载视频图像的增强算法,用于对车载HUD系统进行改进.首先,将退化图像的每个像素RGB三个通道的最大值作为图像初始照射光的估计;其次,针对初始估计的照射光不准确的问题,采用基于加权L1正则化的方法对已估计的照射光进行优化,然后根据Retinex理论将原图像与优化后照射光的比值作为增强后的图像;最后,针对上一步增强后图像出现的高亮区域的过增强以及暗区域的噪声被放大的现象,设置相应权值将原图像和增强后的图像加权合成为最终的图像.通过与现有主流算法的比较表明,该算法增强后的图像细节清晰,色彩恢复度高,对噪声抑制较好,且算法运行速度较快.
Abstract:
In order to solve the problem that the current vehicle HUD system does not deal with the low illumination road conditions,we propose a new enhanced algorithm for the single nighttime vehicle-mounted video image for improvement of the HUD system. First,the maximum value of three RGB channels of each pixel in the degraded image is set as the initial estimated illumination. Secondly,aiming at the problem of inaccurate initial estimated illumination,the estimated illumination is optimized by weighted L1 regularization,and the ratio of the initial image to the optimized illumination is used as the enhanced image according to the Retinex theory. Finally,the corresponding weights are set to address the over-enhancing of the highlight area and amplifying of the noise in dark area,and the initial image and the enhanced image are composed into the final image according to the weights. The comparison with the existing advanced algorithms shows that the enhanced image processed by the proposed algorithm is clear in details with high color restored degree and low noise,and the algorithm runs faster.
更新日期/Last Update: 2018-06-06