Fog Effects Modeling and Removal for Real-Time Vision Applications

Multiple light scattering, leading to blurring or loss of spatial resolution in dense fog regions, is not considered by previous imaging models. To better understand weather effects on vision and improve present defogging methods, this paper describes a new physics-based model, which can explain multiple scattering effects as well as the airlight and attenuation of reflections under fog conditions. Based on the model, an approach for estimating blurring effects of multiple scattering is proposed, using rough depth dependent weighted sum of multi-scale image convolutions. Then, we present a fast single image-based fog removal algorithm, in which blurring effects and airlight are estimated simultaneously with local minimum filter and two-scale Gaussian filters. Experimental results show that the algorithm can make impressive performance with real time implementation.