基于邻域统计信息的红外与可见光图像融合  

Fusion algorithm for infrared and visible images based on neighbor statistic information

作　　者：孙新德[1] 薄树奎[1] 

机构地区：[1]郑州航空工业管理学院计算机科学与应用系,河南郑州450015

出　　处：《激光与红外》2014年第3期339-342,共4页Laser & Infrared

基　　金：国家自然科学基金项目(No.41001235);航空科学基金项目(No.2011ZC55005);河南省科技公关计划项目(No.132102210545)资助

摘　　要：针对红外图像与可见光图像的融合问题,提出了一种基于邻域统计信息的图像融合新算法。首先对图像进行多尺度分解,得到一系列子带系数,然后针对各子带系数的物理特性,提出了高低频规则不同的图像融合算法。对于图像低频部分,首先定义基于邻域统计信息的目标和场景特征参数,然后设计了加权系数自适应变化的加权平均融合策略;对于图像高频部分,首先定义邻域系数分布特征参数,然后设计了受邻域统计信息调制的系数比较取大融合策略。实验结果表明该算法能够很好地将红外图像与可见光图像进行融合,且融合效果优于其他一些算法。Aiming at the fusion problem of infrared and visible images with the same scene, a novel fusion algorithm based on neighbor statistic information is proposed. Firstly the source images are multi-scale decomposed, then many subband coefficients are obtained. Fusion methods of different fusion rules at high and low frequency are presented according to the physical characteristics of each subband coefficient. For the low frequency subband coefficients, the target and scene parameters based on neighbor statistic information are defined, and a weighted average fusion strategy with weighted coefficient adaptive variation is designed. For the high frequency subband coefficients, the neighborhood distribution characteristics parameter is defined, and a fusion strategy of coefficient comparison with neighbor statistic information modulation is designed. The experimental results show that the proposed algorithm can fuse infrared and visible images well.

关 键 词：图像融合 多尺度分解 邻域统计 融合规则 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]