Medical image fusion based on sparse representation and neighbor energy activity

Medical image fusion has become popular in recent years. The fused image can provide a more objective and comprehensive description of lesions and has significant clinical medical aid potential. In this paper, we propose a novel medical image fusion method based on sparse representation and neighbor energy activity that improves the quality of fused images and preserves key information in the source images, such as details, brightness, and color. The proposed method divides the source image into base and detail layers and adopts sparse representation to fuse the base layers. Further, a neighbor energy activity operator that effectively captures the changing fea-tures in the detail layers is utilized. The fused result is obtained by combining the selective layers. The proposed method is applicable to both grayscale and color image fusion problems. In experiments, ten sets of medical images were used as test images. The images included seven different diseases and one normal cranial image and covered five different fusion types: CT/T2, Gad/T2, PET/T1, PET/T2, and SPECT/T2. Further, it was compared with 11 state-of-the-art fusion algorithms, with six highly recognized metrics used for quantitative evaluation. The experimental results indicated that the proposed method outperformed several of the state-of-the-art methods in visual and objective evaluations. Additionally, in experiments conducted to medically analyze the fused images with eight different lesion conditions in the fused images, the fusion results were found to be practicable for medical assistance in actual clinics.