基于MSRCR的自适应低照度图像增强  

Adaptive low-illumination image enhancement based on MSRCR

作　　者：王利娟 常霞[1] 张伯妍 WANG Lijuan;CHANG Xia;ZHANG Boyan(Ningxia Key Laboratory of Intelligent Information and Big Data Processing,North Minzu University,Yinchuan 750021,China;Institute of Image Processing and Understanding,School of Math and Information Science,North Minzu University,Yinchuan 750021,China)

机构地区：[1]北方民族大学宁夏智能信息与大数据处理重点实验室,宁夏银川750021 [2]北方民族大学数学与信息科学学院图像处理与理解研究所,宁夏银川750021

出　　处：《现代电子技术》2022年第2期155-161,共7页Modern Electronics Technique

基　　金：国家自然科学基金项目(11961001);宁夏自然科学基金项目(2018AAC03126);宁夏高等学校一流学科建设(数学学科)(NXYLXK2017B09);北方民族大学重大专项项目(ZDZX201801);宁夏智能信息与大数据处理重点实验室开放基金(2019KLBD004);北方民族大学2020年研究生创新项目(YCX20092)。

摘　　要：针对多尺度Retinex处理低照度图像出现的“光晕伪影”和色彩泛白现象,文中提出一种基于自适应权重带色彩恢复因子的多尺度Retinex图像增强算法。在HSV颜色空间中先将亮度通道图像分解为Retinex增强层和细节恢复层。在Retinex增强层中,不同尺度参数具有不同的增强效果,根据像素的概率分布,计算明暗不同区域的概率分布函数,获得自适应权重。所提算法有效地克服了尺度参数对亮度信息恢复造成的过增强现象。在细节恢复层中,导向滤波具有更优越的保边去噪特性,故采用导向滤波将图像分解成平滑层和边缘层图像,并利用增益系数增强边缘层信息。最后将自适应权重后Retinex亮度增强层、平滑层和边缘层图像融合重构为增强后的亮度通道图像,并在伽马校正算法中融入自适应调节因子来恢复图像在融合过程中丢失的部分细节和色彩信息。实验数据表明所提算法较其他对比算法具有更明显的优越性。In allusion to the phenomenon of"halo artifact"and color whitening in the multi-scale Retinex processing lowillumination images,a multi-scale Retinex image enhancement algorithm based on adaptive weights with color restoration factors is proposed.In the HSV color space,the luminance channel image is decomposed into Retinex enhancement layer and detail restoration layer.In the Retinex enhancement layer,different scale parameters have different enhancement effects.According to the probability distribution of pixels,the probability distribution function of different areas of light and dark is calculated to obtain the adaptive weights.The algorithm can be used to effectively overcome the over-enhancement phenomenon of the brightness information recovery caused by the scale parameter.In the detail restoration layer,as the guided filtering has better edge-preserving and denoising characteristics,the guided filtering is used to decompose the image into the smooth layer and the edge layer images,and the gain coefficient is used to enhance the edge layer information.Finally,the Retinex brightness enhancement layer,smoothing layer and edge layer images after adaptive weighting are fused and reconstructed into the enhanced brightness channel image,and the adaptive adjustment factor is integrated into the gamma correction algorithm to recover some details and color information lost in the fusion process.The experimental data shows that the proposed algorithm has more obvious advantages than other comparison algorithms.

关 键 词：低照度图像 图像增强 RETINEX理论 MSRCR算法 导向滤波 分层融合 伽马校正 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]