Single image dehazing using elliptic curve scattering model

Images obtained in hazy environments present the phenomenon of low contrast, low saturation, and hue offset. Therefore, the main requirements for a dehazing system are accurate color restoration effects and high visual range. In this paper, an improved atmospheric scattering model based on elliptic curve is proposed to remove haze from a single image. Traditional atmospheric scattering model is improved in the following two aspects. Because of uneven illumination in hazy environment, scene incident light is improved by localized operation to replace the global atmospheric light. Considering multiple scattering effects of reflected light, the scene albedo is refined into two parts, the direct transmission item and the indirect transmission item. An elliptic curve model is established to adaptively estimate weights of the two terms. Thus, the reflected light under different conditions of depth and fog density is finely described. Experimental results show that our method achieves good visual performance. Compared with current methods, the indicator of image visibility and visual contrast increased by 40% and 86.6% on average.