Computer Engineering and Applications ›› 2017, Vol. 53 ›› Issue (21): 174-178.DOI: 10.3778/j.issn.1002-8331.1605-0135

Previous Articles     Next Articles

Detection and correction of backlight images

XU Shaoqiu, YANG Qun, ZHONG Xiaoyun, LUO Dehan   

  1. College of Information Engineering, Guangdong University of Technology, Guangzhou 510000, China
  • Online:2017-11-01 Published:2017-11-15

背光图像的检测与校正

许少秋,杨  群,钟小芸,骆德汉   

  1. 广东工业大学 信息工程学院,广州 510000

Abstract: This paper proposes a new backlight image detection and correction algorithm. The change of the number of foreground details under various Gamma correction is analyzed for the determination of backlight image multi-scale Retinex algorithm is performed on the background and the foreground of the backlight image. The outputs are fused with the original image. The experimental results show that the proposed backlight image detection algorithm outperforms the intensity histogram based method and the YCbCr histogram based method, the proposed backlight image correction algorithm has higher information entropy, definition and contrast than the AHMHE backlight image compensation algorithm and multi-scale Retinex algorithm.

Key words: backlight image detection, backlight image correction, Gamma correction, multi-scale Retinex, image fusion

摘要: 针对传统背光图像检测低准确率问题,提出一种新的背光图像检测和校正算法。通过分析图像在不同Gamma变换下前景细节数目的变化规律来判断输入图像是否存在背光。对背光图像的前景、背景子图像进行多尺寸Retinex图像算法,然后与原图进行融合。根据实验可知,提出的背光图像检测算法相对于传统的亮度直方图分析方法以及YCbCr直方图分析方法具有更高准确率;提出的背光校正算法相对于AHMHE背光补偿算法和多尺度Retinex算法具有较高信息熵、清晰度和对比度。

关键词: 背光图像检测, 背光图像校正, Gamma变换, 多尺度Retinex, 图像融合