Single Image Dehazing of Road Scenes Using Spatially Adaptive Atmospheric Point Spread Function

Image haze removal is essential in autonomous driving as the outdoor images captured during unfavorable weather conditions, such or snow, are affected by poor visibility. Much research has been done to overcome image degradation low contrast and faded color due haze. However, traditional model, a phenomenon neglected that several particles simultaneously involved light acquisition. To address this problem, we propose novel single dehazing method based on spatially adaptive atmospheric point spread function (APSF). We developed module estimates APSF limitations of invariant which used existing algorithms. The key factor estimation road scenes with have different statistical characteristic from common hazy resolution. Furthermore, traffic signs lights estimated generating superpixels prevent halo artifacts around sharp edges images. adopted total variation model regularization functional reduce unnatural may occur deconvolution. haze-free proposed tested whether can enhance performance vision algorithms for driving. experimental results demonstrate outperforms state-of-the-art methods enhancing Moreover, additional experiments demonstrated effectiveness quantitative qualitative comparison