基于FRC框架的红外与可见光图像融合方法  

Infrared and visible image fusion based on FRC algorithm

作　　者：代立杨 刘刚 肖刚[2] 阮俊瑾 朱静莲 DAI Li-yang;LIU Gang;XIAO Gang;RUAN Jun-jin;ZHU Jing-lian(College of Automation Engineering,Shanghai University of Electric Power,Shanghai 200090,China;School of Aeronautics and Astronautics,Shanghai Jiao Tong University,Shanghai 200240,China;Shanghai UHDVISION Technology Co Ltd,Shanghai 200040,China)

机构地区：[1]上海电力大学自动化工程学院,上海200090 [2]上海交通大学航空航天学院,上海200240 [3]上海极清慧视科技有限公司,上海200040

出　　处：《控制与决策》2021年第11期2690-2698,共9页Control and Decision

基　　金：国家重点基础研究发展规划项目(2014CB744903);国家自然科学基金项目(61203224);上海浦江人才计划项目(16PJD028)。

摘　　要：像素级图像融合方案多采用全局统一的融合规则,融合过程易导致区域层级信息丢失.对此,提出一种基于模糊区域特征(FRC)框架的红外与可见光图像融合方法.首先,制定一种模糊区域规则,该规则依据区域特征的模糊性质,对源图像依次施行区域分割、模糊判定和解模糊,全局图像的融合转化为3种不同属性区域的融合;然后,分别对目标区域、过渡区域和背景区域采用期望值最大(EM)、加权平均融合、主成分分析(PCA)的融合策略进行融合;最后,融合后的区域图重构得到最终的融合图像.为验证FRC框架保护区域特征的有效性,还提出一种区域特征一致性的度量指标(RCABF),经主观视觉与客观指标检验,融合图像在保持红外目标显著性的同时,具有清晰的背景细节,优于现有的红外与可见光图像融合方法,宜应用于红外目标的监控与分析.In pixel level image fusion schemes,global and unified fusion rules are often used,which results in the loss of regional level information.In this paper,a fusion method of infrared and visible images based on the frame of fuzzy regional characteristics(FRC)is proposed.First of all,a fuzzy region rule is established,according to the fuzzy characteristics of region features,the rule performs region segmentation,fuzzy judgment and defuzziness on the source image in turn.The fusion of global image is transformed into the fusion of three different attribute regions.Then,the target region,transition region and background region are fused by the fusion strategies of expectation maximization(EM),weighted average fusion and principal component analysis(PCA).Finally,the fused region image is reconstructed to get the final fused image.In order to verify the validity of the FRC framework protection region feature,this paper also proposes a region feature consistency metric(RCABF).Tested by subjective vision and objective indicators,the fusion image has clear background details while maintaining the salience of infrared target,which is superior to the existing fusion methods of infrared and visible image,and is suitable for infrared target monitoring and analysis.

关 键 词：红外与可见光图像融合 模糊区域特征 模糊区域规则 期望值最大 主成分分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]