Infrared Visible Image Fusion

Infrared-visible image fusion aims to combine the complementary information from these two image modalities into a single, high-quality image, improving overall scene understanding. Current research focuses on developing efficient and effective fusion algorithms, employing diverse architectures such as convolutional neural networks (CNNs), transformers, and denoising diffusion models, often incorporating techniques like Retinex theory and attention mechanisms to enhance feature extraction and preservation of color and texture details. This field is significant for applications ranging from autonomous driving and fire rescue to medical imaging and object detection, where improved image quality leads to enhanced performance in downstream tasks. Recent work also emphasizes robustness to adversarial attacks and the development of new evaluation metrics tailored to the specific challenges of infrared-visible image fusion.