双指数变换与多尺度细节凸显的红外图像增强  

Infrared Image Enhancement Based on Double Exponential Transform and Multi-Scale Detail Highlighting

作　　者：蔡美芳[1] 钱雯 万里勇[1,2] CAI Meifang;QIAN Wen;WAN Liyong(School of Information and Artificial Intelligence,Nanchang Institute of Science and Technology,Nanchang Jiangxi 330108,China;Management Science and Engineering Research Center,Jiangxi Normal University,Nanchang Jiangxi 330022,China)

机构地区：[1]南昌工学院信息与人工智能学院,江西南昌330108 [2]江西师范大学管理科学与工程研究中心,江西南昌330022

出　　处：《传感技术学报》2023年第7期1079-1085,共7页Chinese Journal of Sensors and Actuators

基　　金：江西省教育厅科学技术研究项目(GJJ212518,G2202907);江西省高等学院教学改革研究课题(JXJG-20-27-10,JXJG-22-27-4);江西省教育科学“十三五”规划课题(19YB266);南昌工学院科技计划博士专项项目(NGKJ-22-01)。

摘　　要：鉴于现有的红外图像增强方法易产生局部的欠增强和过增强以及不清晰的边缘纹理,提出了一种双指数变换与多尺度细节凸显的红外图像增强方法。根据红外图像边缘纹理的粒度,基于Retinex理论用双边滤波将红外图像分解为大、中和小尺度的细节层和基础层;对基础层进行系数小于1的自适应指数变换,在控制亮区过增强的同时自适宜地提高图像的亮度和对比度,而对大、中和小尺度的细节层分别进行系数大于1的自适应指数变换,多尺度地增强图像的边缘细节。实验结果表明,所提方法具有比部分现有方法更优的增强效果,增强图像的视觉效果较清晰,对应的平均梯度和信息熵均较高,而无参考图像质量指标更小。In view of the fact that the existing infrared image enhancement methods are prone to local under enhancement and over en-hancement,and unclear edge textures,an infrared image enhancement method based on double exponential transform and multi-scale detail highlighting is proposed.According to the granularity of the edge texture of the infrared image,and based on the Retinex theory,the infra-red image is decomposed into large-,medium-and small-scale detail layers and basic layer by bilateral filtering.An adaptive exponential transformation with a coefficient less than one is performed on the basic layer to appropriately enhance the brightness and contrast of the image while controlling the over enhancement of the bright areas,and the large-,medium-and small-scale detail layers are subjected to a-daptive exponential transformation with a coefficient greater than one to enhance the multi-scale edges and details of the image.The exper-imental results show that the proposed method has better enhancement effect than some existing methods,the visual effect of the enhanced image is clearer,and the corresponding average gradient and information entropy are higher,while the NIQE index is smaller.

关 键 词：红外图像增强 双指数变换 多尺度细节凸显 双边滤波 无参考图像质量指标 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]