Deep Learning-Enabled Infrared and Visual Image Fusion

Infrared and visual image fusion is an arduous job because of the different modalities of the infrared and visible images leading to their difference in characteristics. The deep learning models and methods have shown much success and results in this field in recent years as it has the ability to study the data and analyze its complex characteristics and features. This paper gives a thorough review of the deep learning-based models and methods developed recently for infrared and visual image fusion such as CNN-based methods, dual-stream-based networks, attention mechanism-based networks, and GAN-based networks. The workflow and concept of each method have been discussed along with their merits and demerits. Flowcharts have also been given along with the discussed methods for providing an understandable explanation, and a comparison has also been drawn among the methods based on several factors such as accuracy and generalizability. This review gives information on the state-of-the-art in deep learning-based infrared and visual image fusion and proves to be a beneficial resource for research purposes in this field.