All-weather road image enhancement using multicolor content-aware color constancy

This paper proposes a method that enhances the road images in real-time, which is an essential part of advanced driver assistance systems. The proposed method restores distorted colors in road images due to illumination by harnessing the relationship between known traffic signs and detected traffic signs via a traffic sign recognition system. The relationship is represented with Von Kries color constancy model which we aim to estimate and apply to the entire image. The proposed method uses a road traffic sign recognition system that is robust against illumination changes. It uses the difference between the detected color values of the traffic sign and an existing reference color values to obtain the coefficients of the Von Kries color constancy method, which is then applied to correct the road images in real time. Our method runs in real time and we tested the proposed method on various road driving images to show superior image enhancement performance regardless of the weather or time of day, compared to methods based on existing image processing techniques and color constancy method such as white balance.