基于HSI颜色空间暗通道的亮度保真去雾方法  

A Brightness Preservation Image Dehazing Method Based on Dark Channel Priori in HSI Color Space

作　　者：汪燕 高静 周岳钰 黄晓晓 闻军 朱柱 WANG Yan;GAO Jing;ZHOU Yueyu;HUANG Xiaoxiao;WEN Jun*;ZHU Zhu(School of Electronic Engineering and Intelligent Manufacturing,Anqing Normal University,Anqing 246133,China;School of Computer Engineering,Jinling Institute of Technology,Nanjing 211169,China)

机构地区：[1]安庆师范大学电子工程与智能制造学院,安徽安庆246133 [2]金陵科技学院计算机工程学院,江苏南京211169

出　　处：《安庆师范大学学报（自然科学版）》2023年第2期61-67,共7页Journal of Anqing Normal University(Natural Science Edition)

基　　金：国家自然科学基金(61701006);安徽省自然科学基金(1708085QF147,2108085QD182);安徽省高校优秀青年人才支持计划重点项目(gxyqZD2019046);安徽省高校自然科学研究项目(KJ2021A0636);安徽高校研究生科学研究项目(YJS20210514)。

摘　　要：图像块和像素先验是实现图像去雾的两类重要策略,但图像去雾后往往会出现亮度损失、过饱和或颜色失真等现象。本文提出了一种基于HSI颜色空间暗通道的亮度保真去雾方法,将RGB颜色空间中传统暗通道模型拓展到HSI颜色空间,以图像亮度、饱和度、颜色浓度和亮度变化构造融合图像块和像素两种先验策略的暗通道模型。通过引入颜色浓度、亮度变化与透射率的两个线性关系,构造出透射率、颜色浓度、亮度变化的方程组,增加了物理散射成像模型参数的可靠性,提高了图像透射率的精度。此外,人造和自然雾天图像去雾实验结果表明,本文方法不但保持了传统暗通道去雾的原有优势,同时也实现了对亮度和饱和度的保真,使复原结果更加逼近真实场景,也更加符合视觉感知。Patch-wised and pixel-wised priories are efficient image dehazing strategies.However,these methods usually result in the degradation of image brightness,saturation,or even the fading of image color.To overcome this problem,we con-struct a dark channel prior(DCP)model in HSI color space instead of the traditional RGB color space by image brightness,sat-uration,color chroma and brightness variance.For more reliable parameters utilized in the atmospheric scattering model,im-age chroma and brightness variance are brought to atmospheric scattering model.Then a novel transmission function is con-structed,thereby improving the accuracy of the image transmission.Experiments on both artificial and natural haze images manifest that our method not only maintains the advantages of the patch-wise priori,but also achieves better preservation of brightness and saturation.The restoration results are close to the real radiance and consistent to visual perception.

关 键 词：图像去雾 暗通道 HSI颜色空间 颜色浓度 透射率 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]