Visible Thermal

Visible-thermal (VT) image processing focuses on integrating information from visible light and thermal infrared cameras to improve various computer vision tasks. Current research emphasizes accurate registration of misaligned VT image pairs, often employing unsupervised generative adversarial networks (GANs) or vision transformer (ViT)-based spatial transformer networks (STNs) to achieve this alignment without manual intervention. This work is crucial for applications like multiple object tracking (MOT) and improving the accuracy of visible-to-thermal image translation, particularly in medical imaging and biometric identification. The development of large-scale, accurately aligned VT datasets is also a key area of focus, enabling the training and evaluation of more robust algorithms.