Infrared and Visible Image Fusion: Methods, Datasets, Applications, and Prospects

Abstract

Infrared and visible light image fusion combines infrared and visible light images by extracting the main information from each image and fusing it together to provide a more comprehensive image with more features from the two photos. Infrared and visible image fusion has gained popularity in recent years and is increasingly being employed in sectors such as target recognition and tracking, night vision, scene segmentation, and others. In order to provide a concise overview of infrared and visible picture fusion, this paper first explores its historical context before outlining current domestic and international research efforts. Then, conventional approaches for infrared and visible picture fusion, such as the multi-scale decomposition method and the sparse representation method, are thoroughly introduced. The advancement of deep learning in recent years has greatly aided the field of picture fusion. The outcomes of the fusion have a wide range of potential applications due to the neural networks’ strong feature extraction and reconstruction skills. As a result, this research also evaluates deep learning techniques. After that, some common objective evaluation indexes are provided, and the performance evaluation of infrared and visible image fusion is introduced. The common datasets in the areas of infrared and visible image fusion are also sorted out at the same time. Datasets play a significant role in the advancement of infrared and visible image fusion and are an essential component of infrared and visible image fusion testing. The application of infrared and visible image fusion in many domains is then simply studied with practical examples, particularly in developing fields, used to show its application. Finally, the prospect of the current infrared and visible image fusion field is presented, and the full text is summarized.