Visible and Infrared Image Fusion Based on Masked Online Convolutional Dictionary Learning with Frequency Domain Computation

In order to avoid the low contrast of image fusion algorithm based on multi-scale transformation and the influence of image slider on the fusion results in the sparse domain, a new visual and infrared image fusion based on masked online convolutional dictionary learning with frequency domain computation is proposed. First, a dictionary filter is obtained by using an online convolutional dictionary learning algorithm with frequency domain computation. Then sparse coefficients are obtained by using convolutional basis pursuit denoising. Finally, fused images are reconstructed. Five classical fusion algorithms are applied to three representative groups of infrared and visible images to prove the advantages of the proposed algorithm. Compared with the CVT-SR-4-based method, NMI, QTE, and QNCIE increased averagely by 83.16%, 15.04%, and 0.46%, respectively.