Single image defogging based on local extrema and relativity of Gaussian

Abstract

Various atmospheric particles such as fog and haze alter the appearance of a natural scene. Fog may afflict many real-life applications such as detecting target objects, tracking, and visibility. The defogging method not only removes fog from images but also causes an improvement in the increase the scene clarity, boost the visual perception of the image, and preserve the structural features. In the proposed work, an improved defogging method based on the local extrema and Relativity of Gaussian is discussed. Here, we consider the model for atmospheric scattering as the background for fog removal. The local extrema method is tailored in such a way as to determine three pyramid levels to calculate atmospheric veil. Then, a multi-scale detail enhancement with Relativity of Gaussian (RoG) is applied to the restored results to produce the images with better appearance. Several experimental analyses are performed on the proposed algorithm to prove that this method achieves more color restoration and detail preservation which have a greater impact on scene perception. This method also focuses on preserving the edges and structures.