All-In-One Image Restoration in Adverse Weather Conditions with Spatial-Frequency Domain Information

Panfeng Jing, Yudong Liang, Siyu Wang, Pin Gao, Hongqiang An

Abstract: Recently, the collaborative operations of multi-domain unmanned vehicle systems have garnered significant attention, particularly regarding their visual perception capabilities under various complex weather conditions. Adverse weather such as rain, snow, and fog can severely degrade image quality, thereby affecting the efficiency and safety of unmanned vehicle systems to a certain extent. Therefore, image enhancement algorithms capable of handling multiple complex weather conditions are crucial for improving the visual perception capabilities of unmanned vehicle systems. This paper presents a unified image enhancement algorithm designed to address various complex weather conditions, such as rain, snow, and fog. The algorithm restores images by integrating spatial and frequency domain information. Experimental results fully demonstrate that this method can significantly enhance images under different adverse weather conditions and greatly improve image quality.