Accurate image segmentation is vital for intelligent transportation systems, enabling reliable detection of vehicles, road structures, and pedestrians across diverse conditions. Conventional methods like global thresholding often struggle with challenges such as lighting variations, fog, shadows, and low contrast. This paper introduces a novel hybrid segmentation technique that, for the first time, integrates the interquartile range (IQR)-based contrast normalization with Otsu thresholding for transportation image segmentation. This combination offers a lightweight, interpretable alternative to deep learning methods while adapting effectively to complex visual environments. The method was tested across real-world transportation scenes—urban roads, traffic scenes, crosswalks, and bridges—under varying conditions (day, night, and fog). Quantitative results show F1 score improvements ranging from 3.23 to 193.97%, significantly outperforming traditional global thresholding. Visual analysis highlights clearer edges, reduced noise, and improved object-background separation, supporting its suitability for robust vision systems in autonomous navigation, traffic monitoring, and smart city infrastructure.