基于NSDTCT与稀疏表示的红外和微光图像融合

Fusion of Infrared and Low-light-level Images Based on NSDTCT and Sparse Representation

针对传统融合方法信息缺失较多与连续性较差等问题,提出了一种基于非下采样双树复轮廓波变换(NSDTCT)与稀疏表示的红外和微光图像融合方法。利用NSDTCT进行多尺度分解获得低频成分和高频子带成分,引入稀疏表示理论,构建低频成分和高频成分的融合模型,将图像融合分别转化为对应稀疏编码的融合,低频和高频成分稀疏表示系数分别根据加权平均和多方向对比度准则进行融合。进行对比实验,选用平均梯度(AG)、标准差(STD)、互信息(MI)、边缘保持度(Q^(AB/F))、结构相似度(SSIM)这5种客观指标,结果较传统方法分别提升2.1%、3.3%、16.1%、6.6%、8.5%以上,表明该方法能够有效保留红外微光图像信息,提高融合图像连续性和成像质量。

Aiming at some shortcomings of the traditional fusion methods,such as lack of information and poor continuity,this paper proposes a new infrared and low-light-level image fusion method based on non-subsampled dual-tree complex contourlet transform(NSDTCT)and sparse representation.Firstly,the NSDTCT multi-scale decomposition was used to obtain low frequency components and high frequency subband components.Secondly,the sparse representation theory was introduced to build the fusion model of low frequency components and high frequency components so as to respectively convert image fusions to fusions of the corresponding sparse coding,with the coefficients of the sparse representations of the low frequency and high frequency component merged respectively according to the rule of the weighted average and multiple direction contrast to merge.Finally,the comparative experiments were conducted with the following five objective indexes selected for comparison,such as average gradient(AG),standard deviation(STD),mutual information(MI),edge retention(Q^(AB/F))and structural similarity(SSIM).As the result,the five indexes,compared with the traditional methods,are improved by 2.1%,3.3%,16.1%,6.6%and 8.5%respectively,indicating that the proposed method can effectively retain the information of infrared and low-light-level images and improve the continuity and image quality of the fused images.