改进Retinex的低照度图像增强研究  

Research on improved Retinex low-illumination image enhancement

作　　者：张平[1] 孙林[1] 解斐斐[1] 田亚楠 赵旭东 ZHANG Ping;SUN Lin;XIE Feifei;TIAN Yanan;ZHAO Xudong(College of Geomatics and Spatial Information,Shandong University of Science and Technology,Qingdao Shandong 266590,China)

机构地区：[1]山东科技大学测绘与空间信息学院,山东青岛266590

出　　处：《激光杂志》2023年第9期91-97,共7页Laser Journal

基　　金：科学技术部高端外国专家引进计划(No.G2021025006L)。

摘　　要：针对低照度图像在增强过程中出现的色彩失真、亮度不均、细节模糊等问题,提出一种改进Retinex低照度图像增强算法。算法基于HSI色彩空间,首先利用Retinex模型将亮度分量I分解为L成分和R成分;采用对比度受限自适应直方图均衡化拉伸L成分,提升图像对比度,基于均方差量化R成分,增强图像整体亮度;并利用均值滤波提取图像细节层,提升图像细节信息;根据I分量与S分量的相关性,提出一种饱和度分量S的自适应拉伸函数,调整图像色彩,最后将增强图像转回RGB色彩空间。实验结果表明,所提算法图像亮度增强适宜,对比度有效提升,细节清晰,图像色彩舒适自然,符合人眼视觉习惯;峰值信噪比、结构相似性、信息熵分别最高达到18.79 dB、0.92、7.59,均高于对比算法。In order to solve the problems of color distortion,uneven brightness and blurred details information in the process of image enhancement of low illumination images,an improved Retinex low illumination image enhance-ment algorithm is proposed.The algorithm is based on HSI color space.First,the brightness component I is decom-posed into L component and R component using Retinex model.The contrast-limited adaptive histogram equalization is used to stretched L component,improve the image contrast,and quantize the R component based on the mean square error to enhance the overall brightness of the image.The mean filter is used to extract the image detail layer and en-hance the image detail information.Next,according to the correlation between I and S,an adaptive stretching function of S is proposed to adjust the image color,and finally the enhanced image is converted back to the RGB color space.The experimental results show that the algorithm in this paper is suitable for image brightness enhancement,effective contrast enhancement,clear details,comfortable and natural image colors,and conform to human visual habits;The peak signal-to-noise ratio,structural similarity index measure and information entropy reached 18.79 dB,0.92 and 7.59 respectively,which were higher than those of the comparison algorithm.

关 键 词：低照度图像增强 HSI色彩空间 改进Retinex 对比度受限自适应直方图均衡化 饱和度拉伸 

分 类 号：TN249[电子电信—物理电子学]