Single Image Dehazing Based on Weighted Variational Regularized Model

Image dehazing is of great significance in computer vision and other fields. The performance mainly relies on the precise computation transmission map. However, existing map still does not work well sky area easily influenced by noise. Hence, dark channel prior (DCP) luminance model are used to estimate coarse this work, which can deal with problem estimation area. Then a novel weighted variational regularization proposed refine transmission. Specifically, simultaneously transmittance restore clear images, yielding haze-free image. More importantly, preserve important image details suppress noise process. In addition, new Gaussian Adaptive Weighted function defined smooth contextual areas while preserving depth discontinuity edges. Experiments real-world synthetic images illustrate that our method has rival advantage state-of-art algorithms different hazy environments.