Haze Density Estimation and Dark Channel Prior Based Image Defogging

Abstract

Among various de-fog algorithms, dark channel prior de-fog algorithm is one of simple and effective dehazing algorithm. The disadvantages of dark channel prior include that there is a certain degree of color distortion in bright areas such as sky and water surface. Aiming at the problem, improved algorithm based on reference-less prediction of perceptual fog density model, Fog Aware Density Evaluator (FADE) is introduced to get more exact estimation of atmospheric light A and medium transmission in the bright areas to avoid the color shift in the sky region. Fast guided filtering is also used in this paper to refine the medium transmission. The result of experiment shows that there is no serious color distortion problem in the sky region of the restored image obtained by the improved algorithm proposed in this paper and the algorithm is more effective than dark channel prior de-fog algorithm.