"Infrared-Visional Dehaze Dataset"

"Current dehazing research predominantly relies on synthetic haze datasets or single visual images. These datasets fail to accurately represent the physical penetration properties of thermal imaging. They also limit the exploration of complementary advantages offered by multi-source fusion in complex environments. Furthermore, existing infrared-visible datasets focus primarily on low-light scenarios rather than hazy conditions. To address this gap, a specialized visible-infrared dataset for hazy scenes is constructed. The dataset contains 300 registered infrared-visible image pairs. It aims to provide high-quality data support for validating the generalization performance of our method and facilitate future research on multi-modal fusion dehazing."