基于多特征的红外与可见光图像融合  

Fusion of infrared and visible images based on multi-features

作　　者：杨桄[1] 童涛[1] 陆松岩[1] 李紫阳[2] 郑悦[1] 

机构地区：[1]空军航空大学 [2]中国人民解放军93010部队

出　　处：《光学精密工程》2014年第2期489-496,共8页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.40901096)

摘　　要：针对传统图像融合方法易导致融合图像整体对比度低及细节反差小的问题,提出一种多特征加权多分辨率图像融合方法。首先,对多尺度分解后的低频系数进行边缘特征、平均梯度特征的提取,同时对高频系数进行相关信号强度比特征的提取。然后,通过边缘特征级融合指导像素级图像融合得到高频系数;针对合成模块中简单加权法易引起边缘或纹理局部模糊的问题,提出分两种情况分别合成同一位置的多尺度分解系数。最后,通过平均梯度特征自适应加权得到融合图像的低频系数,并对低频和高频系数进行多尺度逆变换得到融合图像。实验表明,本文方法的融合性能优于经典的融合方法,其融合质量评价指标中的标准差、空间频率、信息熵和平均梯度分别提高了15.12%、4.30%、6.15%和3.44%。Abstract：In allusion to the lower overall contrast and smaller detail contrast of a fused image from the conventional fusion methods,an effective multi-feature weighted multi-resolution image fusion algorithm was proposed,Firstly,the edge features and average gradient features were extracted from a low frequency coefficient after multi scalar decompose,while the correlated signal intensity ratio feature was extracted from a high frequency coefficient.Then,the high frequency coefficient of the fused image was obtained from the pixel-level weighted average image fusion conducted by the edge feature fusion.Furthermore,a novel combination map was proposed to process the frequency coefficient from the same place with two patterns to solve the problem that the simple weighted method is not effective for retaining the edge and texture information.Finally,the low frequency coefficient of the fused image was obtained by adaptive weighted method based on regional average gradient and the target image was obtained by inversing multi-scale transformation for low frequency and high frequency coefficients.The experiments on fusing infrared and visible images show that the proposed algorithm is better than the classical methods.And the fusion quality indexes,such as standard deviation,spatial frequency,information entropy and average gradient have increased by 15.12%,4.30%,6.15% and 3.44 %,respectively.

关 键 词：图像融合 红外图像 可见光图像 边缘特征 相关信号强度比 平均梯度 特征级融合 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]