基于多尺度SSIF和Gamma变换的低照度图像增强算法  

Low-light image enhancement algorithm based on multi-scale SSIF and Gamma transformation

作　　者：梁杰琛 张鸿铸 魏宗寿 李鹏[3] LIANG Jiechen;ZHANG Hongzhu;WEI Zongshou;LI Peng(Key Lab of Opt-Electronic Technology and Intelligent Control of Ministry of Education,Lanzhou Jiaotong University,Lanzhou 730070,China;Key Laboratory of Plateau Traffic Information Engineering and Control of Gansu Province,Lanzhou Jiaotong University,Lanzhou 730070,China;School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学光电技术与智能控制教育部重点实验室,兰州730070 [2]兰州交通大学甘肃省高原交通信息工程及控制重点实验室,兰州730070 [3]兰州交通大学自动化与电气工程学院,兰州730070

出　　处：《北京交通大学学报》2024年第5期107-117,共11页JOURNAL OF BEIJING JIAOTONG UNIVERSITY

基　　金：甘肃省科技计划项目(21ZD4WA018,22YF7GA140,18JR3RA131);国铁集团科技计划项目(N2022G064);甘肃省教育厅科技项目(2017-A24)。

摘　　要：针对夜间场景下低照度图像整体亮度不足、边缘难以辨识与色彩失真等问题,在HSV色彩空间的基础上,提出一种基于多尺度自引导锐化-平滑图像滤波(Sharpening-Smoothing Image Fil⁃ter,SSIF)的低照度图像增强方法.首先,利用HSV空间色彩亮度分离的特性,对V分量使用多尺度自引导锐化-平滑图像滤波,准确估计光照分量进而求得精确的反射分量.其次,针对光照分量分布不均的问题,提出一种二维自适应伽马变换算法并通过大量对比选取最佳参数,对较暗区域亮度进行拉伸,同时抑制较亮区域的亮度,使整体图像光照更加均匀,图像亮度更符合人眼视觉.再次,针对反射分量存在部分边缘模糊与噪声的问题,提出多尺度钝化掩蔽算法,在抑制噪声的同时能够有效增强图像细节信息,提升整体图像动态范围.最后,对S分量使用自适应饱和度增强算法,将增强后的S分量、V分量与保持不变的H分量合并转到RGB图像,并与带色彩恢复的多尺度视网膜增强算法(Multi-Scale Retinex with Color Restoration,MSRCR)中的色彩恢复因子结合得到最终增强图像.实验结果表明:所提低照度图像增强算法的基于精细自然场景统计的图像质量盲评价指标和平均梯度较其他对比算法分别提高了14.62%、32.10%,不仅能够有效地解决图像亮度分布不均问题,而且能够提高图像轮廓细节的丰富程度和对比度,整体效果优于其他对比算法.To address the challenges of insufficient overall brightness,edge indistinctness,and color distortion in low-light images,this study proposes a low-light image enhancement method based on a multi-scale self-guided Sharpening-Smoothing Image Filter(SSIF)within the HSV color space.First,leveraging the color-luminance separation property of the HSV space,the multi-scale selfguided SSIF is applied to the V component to accurately estimate the illumination component and subsequently extract a precise reflection component.Second,to mitigate the issue of uneven illumination distribution,a two-dimensional adaptive gamma transform algorithm is proposed.Optimal parameters are determined through extensive comparisons,allowing the algorithm to enhance the brightness of darker regions while suppressing the brightness of lighter regions.This results in more uniform image illumination and brightness that aligns with human visual perception.Third,to address edge blurring and noise in the reflection component,a multi-scale passivation masking algorithm is developed,effectively enhancing image details while suppressing noise and improving the overall dynamic range of the image.Finally,an adaptive saturation enhancement algorithm is applied to the S component.The enhanced S and V components are then merged with the unchanged H component and transformed into the RGB image.This output is further refined by incorporating the color restoration factor from the Multi-Scale Retinex with Color Restoration(MSRCR)algorithm,producing the final enhanced image.Experimental results show that the proposed algorithm achieves improvements of 14.62%and 32.10%in noval blind image quality assessment based on fine natural scene statistics and average gradient,respectively,compared to other algorithms.The proposed method not only effectively addresses uneven brightness distribution but also enhances image contour details and contrast,outperforming existing approaches.

关 键 词：信号与信息处理 HSV空间 锐化-平滑图像滤波 多尺度钝化掩蔽算法 二维自适应伽马变换 自适应饱和度增强 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]