基于双MRF模型的夜间图像增强方法  

Night color image enhancement based on dual MRF model

作　　者：赵宏宇[1] 肖创柏[1] 禹晶[1] ZHAO Hongyu XIAO Chuangbai YU Jing(College of Computer Science and Technology, Beijing University of Technology, Beijing 100124, China)

机构地区：[1]北京工业大学计算机学院,北京100124

出　　处：《中南大学学报（自然科学版）》2016年第10期3361-3368,共8页Journal of Central South University:Science and Technology

基　　金：北京市教委科技发展重点项目(KZ01210005007);北京市教育委员会科技发展计划项目(KM201310005020)~~

摘　　要：为了获得更理想的夜间可视效果,通过分析Retinex算法在图像增强时存在的问题,提出一种基于双马尔科夫随机场(MRF)模型的单幅夜间图像增强算法。该算法首先在HSV颜色空间下构造边缘保持的Gaussian-MRF模型对照度分量进行估计,根据Retinex原理获得仅包含物体本身特性的反射分量,并通过增益补偿方法对亮度进行恢复与校正,然后构造Huber-MRF模型对增强结果进行优化,经过颜色空间转换后,最终实现夜间图像的增强。研究结果表明:本文算法增强效果显著,能够有效地凸显边缘细节信息,恢复图像的真实颜色,抑制暗区域噪声,削弱"光晕伪影"的影响,改善夜间图像质量。Through analyzing the errors introduced by Retinex algorithms, a novel enhancement method based on dual Markov random field model（MRF） was proposed to improve single night color image visibility. Firstly, an edge-preserving Gaussian-MRF model to estimate illumination component in HSV color space was constructed. The reflection component obtained by Retinex principle was corrected through gain/offset method. Then, an effective Huber-MRF model to optimize the result was constructed. Finally, after the color space conversion, the proposed method realized the night color image enhancement. The results show that this method has the characteristics of detail preserving, color restoration, Halo effect elimination and noise suppression, and can enhance night color image visibility robustly.

关 键 词：MRF Retinex原理 图像增强 HSV颜色空间 Huber-MRF模型 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]