Fusion of Infrared and Visible Images based on Spatial-Channel Attentional Mechanism

TL;DR

This paper introduces AMFusionNet, a novel infrared and visible image fusion method that uses multi-scale kernels and attention mechanisms to produce high-quality, information-rich fused images, outperforming existing techniques.

Contribution

The paper presents a new deep learning model with multi-kernel convolution and parallel attention mechanisms, enhancing feature extraction and fusion quality in IVIF tasks.

Findings

Outperforms state-of-the-art algorithms in image quality metrics

Achieves significant improvements on public datasets

Effectively preserves thermal and texture details

Abstract

In the study, we present AMFusionNet, an innovative approach to infrared and visible image fusion (IVIF), harnessing the power of multiple kernel sizes and attention mechanisms. By assimilating thermal details from infrared images with texture features from visible sources, our method produces images enriched with comprehensive information. Distinct from prevailing deep learning methodologies, our model encompasses a fusion mechanism powered by multiple convolutional kernels, facilitating the robust capture of a wide feature spectrum. Notably, we incorporate parallel attention mechanisms to emphasize and retain pivotal target details in the resultant images. Moreover, the integration of the multi-scale structural similarity (MS-SSIM) loss function refines network training, optimizing the model for IVIF task. Experimental results demonstrate that our method outperforms state-of-the-art algorithms in terms of quality and quantity. The performance metrics on publicly available datasets also show significant improvement