Image Fusion Algorithm Based on Frequency Enhancement and Shared Tensor Dictionary Learning

With the rapid advancement of multi-source imaging technology, image fusion has become crucial in image processing. This technique combines data from various modalities to produce rich fused images for deeper analysis. Existing multi-scale methods often overlook intrinsic correlations between image components. To address this, we propose an image fusion algorithm based on frequency enhancement and shared tensor dictionary learning. This approach learns a shared multidimensional tensor dictionary for sparse representation of components across frequencies and employs a frequency compensation guided filter (FCGF) to enhance high and low-frequency information. Experimental results show that our algorithm demonstrates strong generalizability in multimodal medical image fusion and infrared-visible image fusion, generating higher-quality fused images.