A Novel Image Defogging Algorithm Based on Improved Bilateral Filtering

Abstract

As we all know, haze and fog lead to decreasing of differentiation degree and definition in outdoor environment and reduce image processing performance. To recover image with fog, we propose a novel fast defogging approach from a gray or color image. First, we use dark channel statistics prior to illustrate the similar relationship between the dark channel and image with fog. Then, to conquer the shortcomings of the estimation methods of fog concentration, we put forward a new filter algorithm. Smoothing the concentration of mist with bilateral filtering can immensely keep the edge of a sudden mutation in the depth of the image. Using partial median filters to smooth the corners can avoid keeping a certain amount of fog around small objects and keep the edge of the corner of the building, which makes the result more robust.