Back to Search Start Over

MFTCFNet: infrared and visible image fusion network based on multi-layer feature tightly coupled.

Authors :
Hao, Shuai
Li, Tong
Ma, Xu
Li, Tian-Qi
Qi, Tian-Rui
Li, Jia-Hao
Source :
Signal, Image & Video Processing; Nov2024, Vol. 18 Issue 11, p8217-8228, 12p
Publication Year :
2024

Abstract

To address the problems of target edge blur and feature loss when fusing infrared and visible images, a novel image fusion network based on multi-layer feature tightly coupled, called MFTCFNet, is proposed. Owing to the difficulty in feature extraction caused by different imaging mechanisms in infrared and visible images, a multi-scale deep feature extraction module has been designed, which consists of designed deformable convolutional-balanced attention mechanism and gradient residual encoder block. Among them, the deformable convolutional-balanced attention mechanism is mainly used to solve the problem of target edge blur caused by single scale features, while the gradient residual encoder block can effectively reduce energy loss in the feature extraction process. In order to fully preserve the feature information of different scales in original images, we construct a multi-layer feature tightly coupled structure. By skillfully utilizing the cross-transmission characteristics of dual branch networks, the problem of feature loss caused by ignoring the correlation between original images can be effectively solved. Extensive experiments showed that the fusion results of the proposed network have more prominent objectives, clearer scene information and better visual effects. [ABSTRACT FROM AUTHOR]

Details

Language :
English
ISSN :
18631703
Volume :
18
Issue :
11
Database :
Complementary Index
Journal :
Signal, Image & Video Processing
Publication Type :
Academic Journal
Accession number :
179636378
Full Text :
https://doi.org/10.1007/s11760-024-03464-y