Visible and Infrared Image Fusion Based on Online Convolutional Dictionary Learning with Sparse Matrix Computation

To overcome the pseudo-Gibbs effect caused by multi-scale transformation fusion method and the block effect caused by sparse representation fusion method, this paper first learns redundant dictionary filter using online convolutional dictionary learning method with sparse matrix computation. The learned dictionary filters are then applied to convolution sparse representation image fusion. Five infrared and visible images are used to prove the performance of the proposed algorithm. The experimental results show that our algorithm not only objectively obtains a higher evaluation index, but also meets the subjective evaluation of human eyes. Compared with the ASR-based method, the normalized mutual information (NMI), the Tsallis entropy (QTE), and the nonlinear correlation information entropy (QNCIE) are increased by 8.34%, 9.16%, and 0.04% averagely.