A Medical Image Fusion Method Based on Unique Features Guidance and Deep Convolutional Networks

Multi-modal medical image fusion technique plays an increasingly important role in the field of image fusion. In this paper, we propose a novel medical image fusion model based on unique features guidance and deep convolutional networks. In our approaches, firstly, feature information from source images is fully extracted by the convolutional neural network. Secondly, under the unique feature guidance, the unique feature information of fused images is enhanced. At the same time, the joint gradient maps are utilized to preserve more texture details in the fused image when performing image reconstruction. Finally, the fusion image is outputted by the feature reconstruction module of the convolutional neural network. The experimental results demonstrate that the proposed method outperforms the other five classic methods in terms of both visual quality and objective assessment.