改进MSRCR的透明物体逆光图像增强算法  

Improved MSRCR Algorithm for Backlight Image Enhancement of Transparent Objects

作　　者：郭成 胡小平 彭向前 黄泓 彭锷 GUO Cheng;HU Xiaoping;PENG Xiangqian;HUANG Hong;PENG E(School of Mechanical and Electrical Engineering,Hunan University of Science and Technology,Xiangtan Hu’nan 411100,China;Hunan Provincial Key Laboratory of Mechanical Equipment Health Maintenance,Hunan University of Science and Technology,Xiangtan Hu’nan 411100,China)

机构地区：[1]湖南科技大学机电工程学院,湖南湘潭411100 [2]湖南科技大学机械设备健康维护湖南省重点实验室,湖南湘潭411100

出　　处：《传感技术学报》2024年第4期645-651,共7页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(61572185);汽车玻璃外观缺陷检测关键算法研究项目(E57110)。

摘　　要：针对透明物体逆光图像对比度低、可视质量差、部分区域过度曝光、边缘信息模糊等问题,提出了一种改进MSRCR(Multi-Scale Retinex with Color Restoration)的透明物体逆光图像增强算法。首先,在MSRCR算法的基础上添加最小可觉差的倒数作为光照分量的调节因子,解决图像的色偏问题,得到准确清晰的透明物体边缘信息;然后,利用自适应对比度增强算法对原图像进行处理,得到亮度适中,对比度高的图像;最后,将两幅图像按亮度均值比例进行拉普拉斯金字塔融合,并进行线性拉伸。将该算法应用于安瓿瓶视觉尺寸测量,结果表明:改进MSRCR的透明物体逆光图像增强算法,能有效解决MSRCR算法的色偏问题,突显透明物体边缘细节信息,并保留亮度增强效果,将安瓿瓶的尺寸误差由0.35 mm降低到0.1mm,提高了透明物体尺寸测量精度。Targeting at the problems of low contrast,poor visual quality,overexposure of some areas and blurred edge information of transparent object backlight image,an improved MSRCR(Multi-Scale Retinex with Color Restoration)algorithm for transparent object backlight image enhancement is proposed.First of all,in order to solve the problem of color deviation of image,obtain accurate and clear edge information of transparent objects,reciprocal of the just noticeable difference is used as an adjustment factor for the illumination component of the MSRCR algorithm.Then,the adaptive contrast enhancement algorithm is used to process original image,and the image with moderate brightness and high contrast is obtained.Finally,the two images are fused with Laplacian pyramid fusion method accord-ing to the average luminance,and then stretched linearly.The algorithm is applied to vision measurement of ampoule dimensions.The results show that the improved MSRCR’s transparent object backlight image enhancement algorithm can effectively solve the color devia-tion problem of MSRCR algorithm,highlight the edge details of transparent objects,retain the brightness enhancement effect,reduce the error of ampoule dimensions from 0.35 mm to 0.1 mm,and improve the accuracy of transparent object size measurement.

关 键 词：图像增强 MSRCR算法 最小可觉差 图像融合 透明物体 逆光图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]