Abstract: Infrared and visible-light images are widely used in various fields owing to their complementarity. However, because of inadequacies in infrared target extraction, directly synthesizing fused images can lead to distortion and information loss. This paper proposes an infrared and visible light image fusion algorithm based on a fast-rolling guidance filter (FRGF) and an improved genetic algorithm. First, the input infrared and visible-light images were subjected to FRGF multiscale decomposition to obtain the base layer and detail layer images. The optimal threshold is then calculated based on the improved genetic algorithm and Renyi entropy to extract the target area from the infrared image. Finally, the base layer was fused using a comparative matching maximum entropy fusion mechanism, and the detail layer was fused using a modified Laplacian energy fusion method. This algorithm combined the advantages of multiscale decomposition and adaptive threshold segmentation. Experimental results show that the proposed algorithm outperforms various classical fusion algorithms in terms of both subjective and objective evaluation metrics, thereby generating superior fusion results.