Single image haze removal based on luminance weight prior

Abstract

In this paper, we propose a simple and effective luminance weight prior for single image dehazing. This prior is based on the observation that the atmospheric airlight closely relates to luminance of haze-free image. Plenty of statistical experiments validates that, at each pixel, the normalized luminance of input image can represent the portion of global atmospheric light that reaches the camera. The luminance weight prior utilizes the normalized luminance matrix to approach the weight map of the atmospheric airlight. After refining the pixel-wise rough luminance weight map by guided filter, with an adaptive postprocessing method we adjust the range of the luminance weight matrix to handle the dense hazy images. Our experimental results and comparative experiments demonstrate that our physically-based algorithm can get clear and high contrast haze-free images, even for dense hazy images.