SF motivated MRI and PET image fusion by combining PCA and RIM under IHS transform

The fusion of multimodal brain images for a given clinical application is importance. A PET image indicates brain function but has low spatial resolution, while an MRI image shows brain tissue anatomy and contains no functional information. Hence, a perfect fused image should contain both functional information and spatial characteristics with no spatial and color distortions. The intensity-hue-saturation (IHS) transform and retina-inspired model (RIM) fusion technique can preserve more spatial feature and more spectral information content, respectively. Moreover, principal component analysis (PCA) algorithm can extract main feature to minimize redundancy. The proposed algorithm integrates their advantages to improve fused image quality. The experiment demonstrates that the proposed algorithm outperforms conventional fusion methods such as PCA, Brovey transform (BT), RIM, discrete wavelet transform (DWT) in light of visual effect and quantitative evaluation.