An Effective Scale-Aware Edge-Smoothing Weighting Constraint-Based Weighted Guided Image Filter for Single Image Dehazing

This paper proposes a new effective scale-aware edge-smoothing weighting constraint-based weighted guided image filter (ESAESWC-WGIF) for single image dehazing. Edge-weighting constraint incorporated in this method is multi-scale and less sensitive to regularization parameter. It removes halo artifacts and over-smoothing strongly and preserves edge information in both flat and sharp regions more accurately than the guided image filter (GIF) and weighted guided image filter (WGIF). There are three main steps in the proposed method: In the first step, dark channel prior method is applied to hazy input image to estimate atmospheric map and transmission map. In the next step, we refine the initial transmission map using the proposed ESAESWC-WGIF. It removes halo artifacts, over-smoothing effect strongly and preserves edge information in both flat and sharp regions. In the final step, the haze-free image is recovered from the scene radiance. About 3200 images from Fattal, NYU2, D-HAZY, Haze-RD, and O-Haze datasets are used to compare the performance of the proposed filter with the existing image dehazing methods. Experimental results prove that the proposed method is independent of the nature of the input image. Moreover, it produces better visual quality. It is noteworthy that the proposed method is faster than the existing methods for a given resolution of images.