A New Robust Scale-Aware Weighting-Based Effective Edge-Preserving Gradient Domain Guided Image Filter for Single Image Dehazing

Abstract

Haze removal is highly desired in computational photography, image processing and computer vision applications. The outdoor images captured in the haze, fog, and mist conditions still have poor color, contrast, and visibility. Recently, many haze removal algorithms have been proposed to restore these images accurately. However, halo artifacts, over-smoothing, and color distortions still persist in the outcomes. This paper proposes a new robust scale-aware weighting based on an effective edge-preserving filter to firmly remove halo artifacts, over-smoothing, and color distortions. The proposed haze removal approach has four steps: First, a dark channel prior (DCP) method is applied to the hazy input image to estimate the atmospheric- and transmission- maps accurately. In the second step, a new robust scale-aware weighting-based effective edge-preserving gradient-domain guided image filter (RSAW-EEPGDGIF) is proposed to refine the transmission map accurately. The proposed RSAW-EEPGDGIF refines the transmission map by decomposing it into the effective weighted-base layer (EWBL) and effective weighted detail layer (EWDL). A non-linear mapping (NLM) function is employed in the next step to suppress various artifacts and noises and enhance the gradient information in EWDL. Finally, we combine the atmospheric map with EWBL and enhanced EWDL to obtain the dehazed output image. The proposed method removes halo artifacts, over-smoothing, and color distortion strongly and preserves edge information precisely in both flat and sharp regions. The theoretical analysis and experimental results prove that the proposed method removes haze fast and more accurately than the other existing dehaze methods. © 2023, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.