Abstract :

Fog and haze significantly degrade visibility in outdoor traffic surveillance videos, leading to loss of important scene information, reduced detection accuracy, and poor monitoring reliability. This work presents a moving objects based real-time defogging approach that enhances traffic videos while maintaining structural consistency across frames. The proposed system first detects moving vehicles and pedestrians, then selectively applies haze removal on dynamic and relevant regions to reduce computational burden. A combination of atmospheric scattering model estimation, temporal stabilization, and contrast enhancement is utilized to restore video clarity. The method aims to improve visibility, color fidelity, and object boundary sharpness in real-time surveillance environments. The enhanced video supports traffic analysis, safety monitoring, and intelligent transportation applications. Experimental results demonstrate improved clarity and object recognition compared to existing defogg