<p>Existing image fusion approaches typically achieve satisfactory performance under standard environmental conditions, yet the growing demand for intelligent decision-making requires advanced imaging capabilities in more complex environments. To address common challenges, such as low contrast, indistinct backgrounds, and insufficiently prominent salient targets in fused images, this study proposes an image fusion algorithm that integrates improved saliency detection with the Non-Subsampled Contourlet Transform (NSCT). The proposed approach strengthens the infrared saliency representation by integrating the Weighted Least Squares (WLS) filter and the Guided Filter (GF). Their complementary smoothing and edge-preserving behavior facilitates a more accurate Frequency-Tuned (FT) saliency refinement. The refined saliency map is further improved through contrast stretching. Subsequently, NSCT is applied to decompose infrared and visible-light images into low-and high-frequency subbands. For low-frequency components, the fusion strategy combines the enhanced infrared saliency map with a Laplacian-based saliency measure (ISML). For high-frequency components, regional energy maximization is integrated with WLS optimization to preserve detailed visible-light textures. The fused image is recovered by applying the inverse NSCT. The experiments show that this approach not only enhances global contrast and background clarity but also maintains visible-light texture details and highlights infrared target regions more effectively. Moreover, the method consistently outperforms existing approaches in objective metric assessments and downstream target detection tasks.</p>