Light source point cluster selection-based atmospheric light estimation

Abstract

The atmospheric light value is a critical parameter in defogging algorithms that are based on atmospheric scattering models. Any error in the atmospheric light value will impact directly on the accuracy of scattering computation and thus cause chromatic distortions in the restored images. To address this problem, this paper proposes a method that relies on clustering statistics to estimate the atmospheric light value. It starts by selecting in the original image some potential atmospheric light source points, which are grouped into point clusters using a clustering technique. From these clusters, several clusters containing candidate atmospheric light source points are selected; the points are then analyzed statistically, and the cluster containing the most candidate points is used for estimating the atmospheric light value. The mean brightness vector of the candidate atmospheric light points in the chosen point cluster is used as the estimate of the atmospheric light value, and their geometric center in the image is accepted as the location of atmospheric light. The experimental results suggest that this statistical clustering method produces more accurate atmosphere brightness vectors and light source locations. This accuracy translates to, from a subjective perspective, both a more natural defogging effect and improvements in various objective image quality indicators.