Exploring the Terrain: An Investigation into Deep Learning-Based Fusion Strategies for Integrating Infrared and Visible Imagery

Abstract

Infrared and visible image fusion technologies influence distinct image features acquired from distinct sensors, preserving complementary information from input images throughout the process of fusion, and utilizing redundant data to enhance the quality of the resulting fused image. Recently, deep learning methods (DL) have been employed by numerous researchers to investigate image fusion, revealing that the application of DL significantly enhances the efficiency of the model and the quality of fusion outcomes. Nevertheless, it is very important to note that DL can be implemented in various branches, and currently, a comprehensive investigation of deep learning-based methods in image fusion is in process.The paper aims to provide an exhaustive review of the evolution of image fusion algorithms grounded in deep learning over the years. Precisely, this paper undertakes a particular exploration of the fusion techniques applied to infrared and visible images through deep learning methodologies. The investigation includes a qualitative and quantitative comparison of extant fusion algorithms using established quality indicators, along with a thorough discussion of diverse fusion approaches. The current research status concerning infrared and visible image fusion is presented, with a forward-looking perspective on potential future directions. This research makes an effort to contribute valuable insights into various image fusion methods developed in recent years, thereby laying a solid foundation for subsequent research goings-on in this domain.