Abstract: 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, (2D)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 are 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.