Multimodal medical image fusion is a computer-aided diagnostic technique designed to integrate effective feature information from different modalities, serving clinical diagnosis and treatment. To address the deficiencies in edge feature preservation and saliency energy perception in existing multimodal medical image fusion methods, this paper proposes a medical image fusion algorithm based on hybrid multi-scale edge preservation and deep image prior-guided illumination saliency decision. First, the truncated Huber filter (THF) is utilized to decompose the source images into a saliency energy layer and a coarse-scale detail layer. Multi-level decomposition latent low-rank representation (MDLatLRR) is then applied to smooth the saliency energy layer and extract fine-scale detail layers. Second, for the base layer, a fusion rule based on illumination map decision guided by deep image prior is used to enhance the visual perception of the fused image. For complex scale edge detail layers, high-frequency nuclear energy mapping is employed to calculate correction weights for fusing the detail layers. Finally, the fusion result is obtained by linearly reconstructing the components. Experiments demonstrate that the proposed method outperforms other state-of-the-art methods in terms of subjective visual quality. Moreover, it achieves average improvements of 6.42%, 16.33%, and 12.58% in the objective metrics QW, QP, and QAB/F, respectively.