基于Retinex和对比度约束的夜间去雾  

Nighttime dehazing based on Retinex and contrast constraint

作　　者：丁雪梅 李为相 毛祥宇 DING Xue-mei;LI Wei-xiang;MAO Xiang-yu(College of Electrical Engineering and Control Science,Nanjing Tech University,Nanjing 211800,China)

机构地区：[1]南京工业大学电气工程与控制科学学院,江苏南京211800

出　　处：《计算机工程与设计》2018年第9期2856-2860,2893,共6页Computer Engineering and Design

基　　金：江苏省"六大人才高峰"基金项目(XXR-012)

摘　　要：针对夜间雾霾天气条件下获取的图像出现细节模糊、颜色失真等问题,提出一种基于Retinex理论和对比度约束的去雾算法。根据Retinex理论求得夜间场景的环境大气光值,通过设置对比度增强与信息损失的约束函数求得透射率估计值,采用导向滤波进行细化,代入夜间雾霾图像光学模型中求得无雾图像,采用一种基于色彩比重的颜色恢复算法对复原图像进行颜色校正。实验结果表明,所提算法能够有效实现夜间有雾图像的去雾处理,恢复的图像细节丰富、色彩自然,在方差、平均梯度、信息熵等指标上均有较高的提升。To address the problems of blurry details and color distortion in the nighttime haze images,a dehazing algorithm based on the theory of Retinex and contrast constraint was presented.The ambient atmospheric light value of night scene was obtained according to Retinex theory.The estimated transmittance value was obtained by the constraint function of contrast enhancement and information loss.The transmittance was refined through the guided filter and was applied into the night haze image optical model,to obtain the restored image.A color restoration algorithm based on color specific gravity was used to adjust the restored image.Experimental results demonstrate that the proposed algorithm can realize dehazing treatment for nighttime haze image effectively.The recovered image has rich details and natural color,many technical specifications of restored images such as va-riance,average gradient and entropy are greatly improved.

关 键 词：图像去雾 RETINEX理论 对比度约束 夜间图像增强 导向滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]