Color image enhancement algorithm based on improved Retinex algorithm

2022 Asia Conference on Algorithms, Computing and Machine Learning (CACML) Pub Date : 2022-03-01 DOI:10.1109/CACML55074.2022.00046

Yuhang Gao, Chuhao Su, Zhaoheng Xu

{"title":"Color image enhancement algorithm based on improved Retinex algorithm","authors":"Yuhang Gao, Chuhao Su, Zhaoheng Xu","doi":"10.1109/CACML55074.2022.00046","DOIUrl":null,"url":null,"abstract":"In order to solve the problem of low expressiveness caused by color distortion and poor saturation when performing image enhancement with classic Retinex algorithm, this paper proposes a color image enhancement algorithm base on the improved Retinex algorithm. In this algorithm, the input image is decomposed into illumination component and reflection component base on Retinex theory first, then logarithmic transfor-mation and Gaussian filtering are performed on the illumination component of HSV color space to approximate the visual system's perception intensity to physical reflectance. Next, the estimated illumination value of scene is used to adjust the multi-scale reflection components of the input image, and to obtain a preliminarily enhanced image. Finally, a color correction factor is introduced into the initial enhanced image to obtain the final enhanced image base on gray world hypothesis. Experimental results show that compared with several classical Retinex algorithms, the proposed algorithm can effectively improve the brightness, contrast and visual information fidelity of the input image without color distortion.","PeriodicalId":137505,"journal":{"name":"2022 Asia Conference on Algorithms, Computing and Machine Learning (CACML)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 Asia Conference on Algorithms, Computing and Machine Learning (CACML)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/CACML55074.2022.00046","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

In order to solve the problem of low expressiveness caused by color distortion and poor saturation when performing image enhancement with classic Retinex algorithm, this paper proposes a color image enhancement algorithm base on the improved Retinex algorithm. In this algorithm, the input image is decomposed into illumination component and reflection component base on Retinex theory first, then logarithmic transfor-mation and Gaussian filtering are performed on the illumination component of HSV color space to approximate the visual system's perception intensity to physical reflectance. Next, the estimated illumination value of scene is used to adjust the multi-scale reflection components of the input image, and to obtain a preliminarily enhanced image. Finally, a color correction factor is introduced into the initial enhanced image to obtain the final enhanced image base on gray world hypothesis. Experimental results show that compared with several classical Retinex algorithms, the proposed algorithm can effectively improve the brightness, contrast and visual information fidelity of the input image without color distortion.

查看原文本刊更多论文

基于改进Retinex算法的彩色图像增强算法

为了解决经典Retinex算法在进行图像增强时由于颜色失真和饱和度差而导致的表现力低的问题，本文提出了一种基于改进Retinex算法的彩色图像增强算法。该算法首先基于Retinex理论将输入图像分解为光照分量和反射分量，然后对HSV色彩空间的光照分量进行对数变换和高斯滤波，使视觉系统的感知强度近似于物理反射率。然后，利用估计的场景照度值对输入图像的多尺度反射分量进行调整，得到初步增强的图像。最后，在初始增强图像中引入色彩校正因子，得到基于灰色世界假设的最终增强图像。实验结果表明，与几种经典的Retinex算法相比，该算法可以有效地提高输入图像的亮度、对比度和视觉信息保真度，且不存在色彩失真。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

2022 Asia Conference on Algorithms, Computing and Machine Learning (CACML)

自引率

0.00%

发文量