Abstract
Advanced driver assistance system (ADAS) can recognize traffic signals, vehicles, pedestrians, and so on all over the vehicle. However, because the ADAS is based on images taken in an outdoor environment, it is susceptible to ambient weather such as fog. Therefore, preprocessing such as de-fog and de-hazing techniques is required to prevent degradation of object recognition performance due to decreased visibility. But, if such a fog removal technique is applied in an environment where there is little or no fog, the visual quality may be deteriorated due to excessive contrast improvement. In this paper, we propose a fog detection algorithm to selectively apply de-fogging method only in the presence of fog. Experimental results show that in the actual images, the proposed algorithm shows an average of more than 97% fog detection accuracy, and improves subjective image quality of existing de-fogging algorithms.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have