一种改进的低照度图像增强算法研究  

Research on an Improved Algorithm for Image Enhancement with Low Illumination

作　　者：刘远仲[1,2] 张强 杨嘉 谭鹤毅[1] 张海波[1,2] 唐天国[1] LIU Yuanzhong;ZHANG Qiang;YANG jia;TAN Heyi;ZHANG Haibo;TANG Tianguo(Nanchong Vocational and Technical College,Nanchong 637000,China;Nanchong Electronic Information Industry Technology Research Institute,Nanchong 637000,China)

机构地区：[1]南充职业技术学院,四川南充637000 [2]南充电子信息产业技术研究院,四川南充637000

出　　处：《四川轻化工大学学报（自然科学版）》2020年第6期46-52,共7页Journal of Sichuan University of Science & Engineering(Natural Science Edition)

基　　金：南充市科技局研发资金项目(19YFZJ0022)。

摘　　要：针对图像在低照度的情况下易出现图像不清晰、对比度不够、细节不能充分显现等不足,提出了一种改进的低照度图像增强算法。首先,对采集到的图像进行光照修正,通过光照修正法将图像分为强光区域与暗光区域,暗光区域与强光区域部分分开后,采用判定函数计算暗光区域像素值,凸显图像细节;其次,为减小计算复杂度,保证算法计算量快速收敛,采用三次快速均值滤波代替高斯滤波处理图像;最后,对处理后的图像进行了主观和客观评价分析,对比该算法的优势。对3种不同场景的城市图像使用不同的算法仿真测试并进行对比,实验表明,该算法相比传统的Retinex算法处理图像,计算复杂度低,计算量小,耗时大为减少;增强后的图像颜色失真小,对比度高,算法能快速有效实时处理低照度图像。In view of the problems that image is easy to be fuzzy,contrast is not enough,and details are not fully revealed.etc in the case of low illumination,this paper puts forward an improved low illumination image enhancement algorithm.First of all,correct the light on the collected image,then the image is divided into light areas and dark areas by the light correction method,after the image region is separated,the dark area pixel values are calculated by using a decision function to highlight the image details;secondly,to reduce the computational complexity and ensure fast convergence of the algorithm computation,this paper uses three times fast average filtering to replace the Gaussian filter to process images;finally,the subjective and objective evaluation and analysis of the processed image are carried out to compare the advantages of the algorithm.In three different scenarios of urban images,using different algorithm simulation tests and comparison,experiments show that compared with the traditional Retinex algorithm to process images,the proposed algorithm has lower computational complexity,less calculation,and time-consuming greatly reduced;the enhanced image has low color distortion and high contrast,the algorithm can quickly and effectively process low illumination image in real-time.

关 键 词：低照度 高斯滤波 RETINEX算法 图像增强 均值滤波 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TP391[自动化与计算机技术—控制科学与工程]