<p>In the field of military target detection, due to the limitations of acquisition conditions in complex battlefield environments, some images still suffer from insufficient clarity even after preprocessing. This not only makes it difficult to provide accurate target feature information required for effective detection, but also significantly reduces real-time computing efficiency due to data redundancy and noise interference. To address this, this paper proposes a military image fusion strategy based on composite decomposition and multi-fuzziness theory, aiming to balance image quality improvement and computational accuracy optimization supported by high-performance computing. Firstly, the two-scale composite method is combined with LatLRR to decompose the source infrared and visible images into low-frequency details, low-frequency significant, and high-frequency images, respectively, to excavate the deep information of the images. Secondly, three fusion strategies based on fuzzy theory, GFVSM weighting, CFEOE adaptive, and improved GSSIFS function, are proposed to fuse the decomposed images in the following manner: the low-frequency detail images are fused using a Gaussian fuzzy function that adjusts the visual saliency map weighting function; the low-frequency salient images are fused using a Cauchy fuzzy function that adjusts the image energy weight; the high-frequency images are fused employing a proposed improved intuitionistic fuzzy set function. Finally, four sets of typical military images in the infrared–visible dataset are used to test the efficacy of our method and to compare it against four mainstream fusion methods, both subjectively and objectively. The comparison experiments results show that the images fused using our method show improvements of 9.56%, 42.12%, 47.46%, 44.94%, and 28.82% in five key performance indices, respectively, over the average values produced by the four mainstream fusion algorithms. The three-dimensional image reveals that the image fused by this method has enhanced rich detail information, prominent target, and obvious edge features. The application experiments results indicate that the military target detection rate of the fused image is improved by 9.04% and 6.87% compared with that of the infrared and visible images before fusion, respectively. This military image fusion strategy effectively retains the key information of the source image thus improving the accuracy of military target detection.</p>