基于WLS的Retinex单幅夜间彩色图像增强方法  

Retinex Algorithm for Night Color Image Enhancement Based on WLS

作　　者：赵宏宇[1] 肖创柏[1] 禹晶[2] 戴岩[1] 

机构地区：[1]北京工业大学计算机学院,北京100124 [2]清华大学电子工程系,北京100084

出　　处：《北京工业大学学报》2014年第3期404-410,共7页Journal of Beijing University of Technology

基　　金：北京市教委科技发展重点项目资助(KZ01210005007);北京市教育委员会科技发展计划项目资助(KM201010005011)

摘　　要：为了提高夜间可视效果,针对单幅夜间彩色图像,作者提出一种基于加权最小二乘(WLS)的Retinex增强算法.首先构造优化的颜色恢复函数,然后在HSV(hue,saturation,value)颜色空间下,根据夜间图像的亮度与梯度特性构建WLS框架对图像照度分量进行估计,获得仅包含物体本身特性的反射分量图像,通过颜色恢复函数与Gamma校正等操作进行颜色恢复与补偿,实现夜间彩色图像增强.实验结果表明:该算法增强效果显著,与其他算法相比,具有消除"光晕伪影"现象、抑制噪声、颜色保真和较好保留边缘细节信息等能力.In order to improve visual effect of single night color image, a Retinex enhancement algorithm based on weighted least squares （WLS） was proposed. Firstly, this algorithm constructed a optimization function to prepare for the color restoration. Secondly, it processed brightness image of HSV （hue, saturation, value） color space using the WLS optimization framework which composed of brightness and gradient characteristics of night image. Then, the scene illumination could be estimated, and reflection image could be obtained. Finally, it finished the night color image enhancement after the color restoration and compensation process. The experimental results show that this algorithm has a good ability of image enhancement. Compared with other algorithms, this algorithm also has the characteristics of color restoration, Halo effect elimination, noise suppression and detail preserving.

关 键 词：WLS RETINEX 图像增强 HSV 夜间彩色图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]