Medical image fusion based on extended difference-of-Gaussians and edge-preserving

Multimodal medical image fusion extracts useful information from different modal medical images and in-tegrates them into one image for a comprehensive and objective lesion description. However, existing methods ignore the simultaneous retention of significant edge and energy information that reflect lesion characteristics in medical images; this affects the application value of medical image fusion in computer aided diagnosis. This paper proposes a novel medical image fusion scheme based on extended difference-of-Gaussians (XDoG) and edge-preserving. A simple yet effective energy-based scheme was developed to generate the fused energy layer, which helped preserve energy. Moreover, the averaging filter was used to generate the detail layers of source images. The fusion of detail layers was considered the combination of significant and non-significant edge in-formation. A rule of the detail layer with a salient edge based on edge extraction operator XDoG was proposed to efficiently detect the salient structure of the significant edges, and a spatial frequency energy operator was developed to detect the gradient and energy of non-significant information. The fused result could be recon-structed by synthesizing the fused energy layer and details of significant and non-significant edges. Experiments demonstrated that the proposed approach outperforms some advanced fusion methods in terms of subjective and objective assessment. The code of this paper is available at https://github.com/JEI981214/FGF-and-X DoG-based.