Multimodal medical image fusion using modified fusion rules and guided filter

Multimodal medical image fusion plays a crucial role in medical diagnostics and treatment. Widely used transform domain based image fusion methods like DWT, CVT, CT, NCST suffer from spatial inconsistency and high complexity. Recently proposed guided filter based spatial domain image fusion techniques are also limited by contrast reduction and halo artifacts. In this paper, the existing guided filter based image fusion scheme is modified by using Gaussian decomposition with local average energy and average gradient based saliency maps for base layer and detail layer respectively. Simulation results show that the modified fusion technique is more effective in preserving contrast and fine details than the existing guided filter based and DWT based fusion scheme.