Research On Multi-mode Image Fusion Algorithm Based On Convolutional Network

Image fusion technology has significant application value in military,medical,and security fields as it can reveal scene information more comprehensively and clearly,benefiting both human vision and machine recognition.In recent years,scholars at home and abroad have conducted extensive and in-depth research on image fusion technology.However,most algorithms only analyze the problem of single-modal image fusion,which may have certain shortcomings when dealing with other modal image fusion problems,resulting in the algorithm’s lack of good universality.This paper aims to effectively fuse multi-modal images and analyzes the advantages and disadvantages of existing image fusion algorithms based on multiscale transformation and convolutional neural network.Based on this,a convolutional neural network-based multi-modal image fusion algorithm is proposed.The main research content and innovative points of this paper include the following aspects:(1)In order to improve the universality of the multi-modal image fusion algorithm,a convolutional feature adaptive fusion framework is designed based on the end-to-end fully convolutional neural network structure and residual module.This method does not require preprocessing such as filter design or specific fusion rules,fully utilizing the generalization ability of convolutional neural networks to improve the universality of image fusion algorithms.(2)In order to reduce the loss of convolutional features of source images in the fusion process,a novel image reconstruction structure combining skip connections and convolutional layers is designed.This structure is beneficial to supplement the image detail information lost during the pooling process and improve the utilization rate of the fusion algorithm for convolutional feature information of the source images.