A Review on Generative Adversarial Network based Infrared and Visible Image Fusion Techniques

The infrared and visible (IRVS) image fusion is an innovative solution that has significantly advanced image processing and computer vision by providing comprehensive frameworks for generating high-quality integrated images. Infrared images represent thermal radiation that distinguishes objects from the background under different lighting and weather, while visible images provide rich texture information. Generative Adversarial Network-based models overcome the limitations of conventional and Convolutional Neural Network-based approaches by eliminating the need for intermediate human intervention and training datasets, enabling seamless fusion. This paper reviews image fusion techniques based on Generative Adversarial Networks (GANs) specifically designed for IRVS image fusion, focusing on their ability to enhance contrast and texture details in the fused outputs. An analysis of widely used IRVS image fusion datasets is presented, highlighting their key characteristics. Finally, the paper discusses some performance evaluation measures that is used to evaluate the quality and effectiveness of different fusion techniques.