NOSMFuse: An infrared and visible image fusion approach based on norm optimization and slime mold architecture

In existing infrared and visible image fusion algorithms, it is usually difficult to maintain a good balance of meaningful information between two source images, which easily leads to the omission of important fractional information in a particular source image. To address this issue, a novel fusion algorithm based on norm optimization and slime mold architecture, called NOSMFuse, is proposed. First, an interactive information decomposition method based on mutually guided image filtering is devised and utilized to obtain the corresponding base and detail layers. Subsequently, the differentiation feature extraction operator is formulated and employed to fuse the base layers. In addition, we design a norm optimization-based fusion strategy for the detail layers and a loss function that considers both the intensity fidelity and the gradient constraint. Finally, to further balance the useful information of the base and detail layers contained in the fusion image, we propose a slime mold architecture based image reconstruction method that generates fusion results through adaptive optimization. The experimental results show that the proposed NOSMFuse is superior to 12 other state-of-art fusion algorithms, both qualitatively and quantitatively.