Single Image Dehazing Using Fixed Points and Nearest-Neighbor Regularization

Natural images captured in bad weather conditions often suffer from poor visibility. Dehazing, the process of removing haze from a single input image or multiple images, is a crucial task in image and video processing, which is quite challenging because the number of freedoms is lager than the number of observations. In this paper, we propose a novel method to reduce the block artifacts and halos for single image dehazing, which replaces the widely used soft matting and contextual regularization. We first find some fixed points in a maximum filter and then apply a Nearest-Neighbor (NN) regularization to recover a smooth transmission map. Compared with the state-of-the-art single image dehazing methods, the experimental results on some typical and challenged images demonstrate that our method can produce a high-quality dehazed image and recover the fine detail information and vivid color from the image haze regions.