LCVR-Net: Dual-Attention Visibility Restoration for Traffic Surveillance Under Dust and Fog Degradation

Adverse atmospheric conditions, particularly dust and fog, substantially degrade the visibility of traffic surveillance imagery, limiting the reliability of intelligent transportation systems and vision-based traffic monitoring applications. To address these limitations, this paper proposes the Lightweight CCTV Visibility Restoration Network (LCVR-Net), an efficient image restoration framework specifically designed for CCTV surveillance under adverse weather conditions. The proposed architecture adopts a lightweight encoder–decoder backbone based on Residual Depthwise Separable Blocks (RDSBs) to achieve effective feature extraction with low computational complexity. Furthermore, a Dual Attention Refinement Module (DARM) is introduced to enhance degradation-aware feature representation for fog restoration, while a lightweight Color Correction Head (CCH) is incorporated to compensate for atmospheric color distortion and improve perceptual image fidelity. To enable task-specific optimization, two model variants are developed for dust and fog restoration, respectively. Qualitative evaluations on real-world surveillance images further provide supporting evidence of the practical applicability of the proposed framework under naturally occurring atmospheric degradation. These results demonstrate that LCVR-Net provides an effective balance between restoration accuracy and computational efficiency, making it well suited for practical deployment in intelligent transportation and traffic surveillance systems.