A Lightweight Image Denoising and Adverse Weather Enhancement Method for Real-Time Visual Perception Services in Autonomous Driving Systems

Real-time visual perception is a critical informatics service in autonomous driving systems, directly supporting downstream decision-making, control, and safety assurance.Under adverse weather conditions such as rain, fog, snow, and low-light environments, images captured by onboard cameras suffer from severe noise, contrast degradation, and detail loss, which significantly reduce the reliability of perception services.Moreover, perception algorithms are typically deployed on vehicle-grade embedded platforms with limited computational resources, making heavyweight denoising and enhancement models unsuitable for real-time operation.To address these challenges, this paper proposes a lightweight image denoising and adverse weather enhancement method designed for real-time autonomous driving perception services.A low-complexity denoising network is developed based on shallow architecture, lightweight convolutional operators, and residual learning.On this basis, targeted enhancement strategies are introduced for rainy, foggy, and low-light scenes, enabling joint optimization of noise suppression and visual enhancement under strict resource constraints.Experimental results on multiple autonomous driving datasets demonstrate that the proposed method improves image quality, reduces inference latency, and enhances the performance of downstream perception tasks such as object detection and semantic segmentation.The results indicate that the proposed approach provides an effective and deployable informatics solution for robust real-time perception services in autonomous driving systems.