Fog Detection and Fog Synthesis for Effective Quantitative Evaluation of Fog-detection-and-removal Algorithms

초록

An advanced driver assistance system (ADAS) can recognize traffic signals, vehicles, pedestrians, and so on, all around the vehicle. However, because the ADAS is based on images taken in an outdoor environment, it is susceptible to ambient weather such as fog. So, preprocessing with de-fogging 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, visual quality may deteriorate due to excessive contrast improvement. And in foggy road environments, typical fog removal algorithms suffer from color distortion. In this paper, we propose a temporal filter-based fog detection algorithm to selectively apply de-fogging only in the presence of fog. We also propose a method to avoid color distortion by detecting the sky region and applying different methods to sky and non-sky regions. There is no publicly available database of foggy images and corresponding fogless images, making it difficult to quantitatively evaluate the performance of fog-detection and fog-removal algorithms. So for road environments, we propose a fog synthesis algorithm based on depth information and temporal filtering. Experimental results show that in the actual images, the proposed algorithm shows an average of more than 97% fog detection accuracy, and improves the subjective image quality of existing de-fogging algorithms. In addition, the proposed algorithm offers fast computation time of less than 0.1 ms per frame. We also show that by using the proposed fog synthesis technique, fog removal algorithms can be evaluated by quantitative evaluation metrics such as peak signal-to-noise ratio (PSNR). Copyrights © 2018 The Institute of Electronics and Information Engineers

키워드