基于CSR和能量特征的红外与可见光图像融合  

Infrared and visible image fusion based on CSR and energy features

作　　者：王昭 杜庆治[1] 龙华[1] 邵玉斌[1] 彭艺[1] WANG Zhao;DU Qing-zhi;LONG Hua;SHAO Yu-bin;PENG Yi(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500

出　　处：《激光与红外》2021年第8期1088-1096,共9页Laser & Infrared

基　　金：国家自然科学基金项目(No.61761025,No.81860318)资助。

摘　　要：传统稀疏表示(SR)分块处理策略降低了图像连续性,使得特征信息损失严重。因此,提出了基于卷积稀疏表示(CSR)和能量特征的红外与可见光图像融合算法。该算法将非下采样轮廓波变换(NSCT)域低频子带分解成低频基础分量和细节特征分量,使用局部拉普拉斯能量法(LLE)和卷积稀疏表示分别进行融合,获得低频子带融合图像。同时,根据底层视觉特征构建新活性度量方法来融合高频子带,最后对高、低频部分进行NSCT反变换重建。实验结果表明:该算法有效结合了源图像的边缘纹理信息,在主观和客观评价上皆优于现有的大部分算法。The traditional sparse representation(SR)block-processing strategy reduces the continuity of the image,which results in a serious loss of feature information.Therefore,an infrared and visible image fusion algorithm based on convolution sparse representation(CSR)and energy features is proposed.In this algorithm,the low-frequency subband of the non-subsampled contoured transform(NSCT)domain is decomposed into the low-frequency basic component and the detailed feature component,local Laplace energy method(LLE)and convolutional sparse representation are used for fusion respectively to obtain the low-frequency subband fusion image.Meanwhile,a new activity measurement method is constructed according to the underlying visual features to fuse the high-frequency subband.Finally,the low-frequency and high-frequency parts are reconstructed by NSCT inverse transformation.The experimental results show that this algorithm effectively combines the edge texture information of source images,and is superior to most existing algorithms in subjective and objective evaluation.

关 键 词：图像融合 非下采样轮廓波变换 局部拉普拉斯能量法 卷积稀疏表示 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]