基于加权结构感知和雾天退化模型的低照度图像增强方法  

A low-illumination image enhancement method based on structure-aware weighting function and haze-degradation model

作　　者：姚斌[1] 魏玄颖 韩振 何立风 YAO Bin;WEI Xuan-ying;HAN Zhen;HE Li-feng(School of Electronic Information and Artificial Intelligence, Shaanxi University of Science & Technology, Xi′an 710021, China;Faculty of Information Science and Technology, Aichi Prefectural University, Aichi 4801198, Japan)

机构地区：[1]陕西科技大学电子信息与人工智能学院,陕西西安710021 [2]日本爱知县立大学信息科学学院,日本爱知县长久手市4801198

出　　处：《陕西科技大学学报》2022年第4期165-170,190,共7页Journal of Shaanxi University of Science & Technology

基　　金：国家自然科学基金项目(61603234,61971272);陕西科技大学博士科研启动基金项目(2020BJ-18)。

摘　　要：基于反转的低照度图像与雾天图像的相似性,结合物理模型和数学优化理论,提出了一种改进的基于去雾模型的低照度图像增强方法.该方法通过设计一种具有结构感知性的权重函数,构建了一个适用于雾天退化模型的透射率目标函数,从而对反转的低照度图像进行去“雾”.与经典低照度图像增强方法相比,该方法增强的图像颜色自然,细节保持良好,同时在信息熵、平均梯度、自然图像质量评价指标表现较好.According to the similarity between the inverted low illumination image and hazy image,an improved low-illumination image enhancement method based on dehazing model is proposed in this paper,which combines physical model and optimization theory.This work designs a structure-aware weighting function and constructs an optimal transmission function suitable for dehazing model,thus,the inverted low-illumination image is″dehazed″.Compared with some classical low-illumination enhancement methods,images enhanced by this method have natural colors and vivid details.Besides,this method obtains better image quality assessment results in metrics such as IE,AG,and NIQE.

关 键 词：低照度图像 去雾方法 边界约束 透射率图 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]