Video De-Fogging based on Improved Color Attenuation Prior

  • Isha Kansal, Renu Popli, Geetanjali Kapoor


This papercomes up with a fog removal technique for fog degraded videos based upon improved Color Attenuation Prior called as CAP. CAP uses a machine learning techniqueforfinding the depth or transmission map of a fog degraded image. Thismap is further utilized for finding the unknown parameters of physical foggy image degradation model.  Window based local minimum operation is applied in an efficient way so as to speed up the process and a fast Gradient Domain Guided Filter called as GDGF is used with CAP based transmission map for making the de-fogging process more faster and visually appealing. The edge preservation property of GDGF produces better quality de-fogged images. Since the image de-fogging techniques can even be applied on foggy videos but the task involves some extra internal steps due to temporal nature of videos. In videos,large number of frames are captured per second, it is more likely to have number of frames adjacent to each other are similar or correlated. By utilizing this property, in this work the de-fogging speed is further enhanced in the proposed work.