关于混合多距离图像运动目标Retinex增强研究  

Research on Retinex Enhancement of Moving Target in Mixed Multi Range Image

作　　者：程学军 潘红改 王建平[2] CHENG Xue-jun;PAN Hong-gai;WANG Jian-ping(Henan University of Technology,Luohe Institute of Technology,Luohe Henan 462000,China;School of Information Engineering,Henan Institute of Science and Technology,Xinxiang Hean 453003,China)

机构地区：[1]河南工业大学漯河工学院,河南漯河462000 [2]河南科技学院信息工程学院,河南新乡453003

出　　处：《计算机仿真》2021年第12期105-108,290,共5页Computer Simulation

基　　金：河南省科技攻关计划项目(212102210422);河南省高等学校重点科研项目(20A520002);河南省高等学校青年骨干教师培养计划项目(2019GGJS172)。

摘　　要：当前的图像识别主要针对清晰的近距离目标,随着混合多距离图像应用场景需求的增加,现有图像处理算法对混合多距离图像中运动目标的亮度、清晰度和分辨率变化表现出明显的性能缺陷。针对上述问题,提出了一种Retinex图像增强方法。利用Retinex对图像成分进行分解,得到图像的边缘和细节特征,设计了简化传函来降低变量个数和控制复杂度,并在反射分量中引入包含距离、亮度和多尺度的灰度系数来避免灰度突变。为防止出现过增强和反射率波动,针对细节设计了可变增强因子,针对亮度引入多尺度伽马变换。为避免增强图像失真,利用细节、对比度、亮度等图像特征参数构建损失函数。仿真从主观与客观两个方面进行验证,结果显示所提方法能够对混合多距离图像中的运动目标进行有效提取,同时保持较高的PSNR和SSIM指标,说明方法能够适应图像的亮度、清晰度和分辨率变化,提高运动目标的增强性能。The current image recognition mainly aims at clear close-range targets. With the increasing demand of hybrid multi-distance image application scenarios, the existing image processing algorithms have obvious performance defects on the brightness, clarity, and resolution changes of moving objects in Mixed Multi-range images. Based on this problem, a Retinex image enhancement method is proposed. Retinex is used to decompose the image components to obtain the edge and detail features of the image. In this process, a simplified transfer function is designed to reduce the number of variables and control complexity, and a gray coefficient including distance, brightness and multi-scale is introduced into the reflection component to avoid gray mutation. In order to avoid over enhancement and reflectivity fluctuation, a variable enhancement factor is designed for details, and a multi-scale gamma transform is introduced for brightness. In addition, in order to avoid the distortion of the enhanced image, the loss function is constructed by using the image feature parameters such as detail, contrast, and brightness. Simulation results show that the proposed method can effectively extract moving objects from Mixed Multi-range images, while maintaining high PSNR and SSIM indexes. It shows that the method can adapt to the changes of image brightness, clarity, and resolution, and improve the enhancement performance of moving objects.

关 键 词：混合多距离图像 多尺度变换 损失函数 图像增强 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]