Medical image fusion algorithm based on L0 gradient minimization for CT and MRI

In this paper, a novel medical image fusion method based on L₀ Gradient Minimization for CT and MRI is proposed. Compared with traditional algorithms, the proposed method performs well in preserving bones structures from CT and sustaining the soft tissue detail from MRI. It’s worth mentioning that both the proposed low- and high-frequency fusion rules have the capability of generating appropriate weight maps according to the characteristics of CT and MRI images. The fusion algorithm using L₀ Gradient Minimization mainly comprises of four steps: First, source images are decomposed into multi-scale representations via L₀ Gradient Minimization. Second, we propose a low-frequency fusion rule based on local energy and Gaussian filters, which can generate the fused base layer in accord with the basic principle of human beings’ visual system. Third, high-frequency sub-bands are fused by utilizing saliency detection rule based on texture extraction, which generates the satisfying maps according to the degree of significance. Finally, we get the fused result according to the image reconstruction. The proposed algorithm is compared with nine advanced fusion methods and shows superior performance in whether subjective or objective evaluations.