基于（2D）^2-KL变换的小波域图像融合算法

Image Fusion Algorithm Based on the （2D）^2-KL Transform in Wavelet Domain

为使融合后的图像在尽可能保留源图像细节信息的同时,还能够有效提高源图像的对比度,提出基于(2D)2-KL((2D)2-Karhunen-Loeve)变换的小波域图像融合算法.首先用(2D)2-KL变换直接对图像信息进行分析,并构建协方差阵,提取图像的重要特征,然后将其主要特征输入到小波域中.在此基础上,对小波变换分解得到各子带系数,用一定的融合策略进行融合.低频子带含有图像的轮廓信息,引入加权因子指导低频子带系数进行融合.实验结果表明,提出的算法有效提高了图像的对比度,并且很好地保留了图像的细节信息,无论在视觉角度上,还是在各种客观性能评价上都比其它传统方法取得了更佳的融合效果.

An image fusion algorithm based on the （2D）^2-KL transform in a wavelet domain is proposed, aiming at reserving the details of the original images after image fusion and improving the contrast of the original images. First, （2 D）^2-KL transform is used to analyze the information directly and the covariance matrix is constructed to obtain the significant features. Then, the significant features are transmitted into the wavelet domain. On this basis, certain fusion strategies arc utilized for each subband coefficient of the wavelet decomposition. The low frequency subband contains the contour information of the image, and therefore, the weighting factors are employed to fuse the low-frequency coefficient. Finally, experiments show that the proposed method can not only effectively improve the contrast of the image, but also retain the image details. By comparing human perceptions and conducting various objective performance evaluations, the proposed method is shown to perform better than conventional approaches.