Infrared and visible image fusion integrates complementary data from multi-sensor imagery into a single, information-rich composite, enhancing scene understanding across diverse applications. Despite recent advances in machine learning-driven fusion, challenges such as artifacts, blurred features, and poorly enhanced critical regions persist. We propose a novel Fusion Generative Adversarial Network (FGAN) that synergistically combines a fuzzy logic- based generator with a support vector machine (SVM)-powered discriminator. The generator leverages a Mamdani-type fuzzy logic system, optimized via the Harris Hawks Optimization (HHO) algorithm, targeting entropy, PSNR, and SSIM metrics to refine fusion quality. Concurrently, the discriminator employs the Frechet Inception Distance (FID) to robustly distinguish real and synthetic images. Evaluated on the TNO dataset using MATLAB, our FGAN delivers superior subjective visual quality and objective performance, outperforming state-of-the-art methods and setting a new benchmark for image fusion.
Zarimeidani et al. (Thu,) studied this question.