CT and MR Image Fusion Based on Adaptive Structure Decomposition

Abstract

Computed tomography (CT) has an excellent performance in detecting dense structure, such as bones and implants, while magnetic resonance (MR) provides high-resolution information for soft issues. To obtain sufficient and accurate information for diagnosis, we propose a CT and MR image fusion method via adaptive structure decomposition to combine the complementary information. First, on the basis of different scales of issues, we adaptively decompose the source images into sub-bands (bands of small, middle, and large issues) by a spectral total variation method. Second, based on the interpretability of sub-bands, for the small scale and middle scale of issues, we extract the edge information from the sub-bands and design the fusion weight by the local edge energy. And for the large scale of issues, we design the fusion weight by the local intensity energy. Third, we reconstruct the fused image. The experimental results demonstrate the superiority of the proposed method on both subjective and objective assessments.