A High-Efficiency and High-Speed Gain Intervention Refinement Filter for Haze Removal

The dark channel prior has been considered to be an efficient dehazing technique in recent years. However, its operation causes annoying halo effects. To solve this problem, many former imaging filters have been proposed and combined with the dark channel prior operation. However, these filters inevitably induce enormous computational burden while the dehazing effect of the dark channel prior still has room for improvement. To cope with this, a high-speed refinement method based on the gain intervention is proposed and combined with the dark channel prior to solve the aforementioned problems. As demonstrated in our experiments, the proposed filter integrated into the dark channel prior yields not only higher processing speeds but also superior recovery effects than can previous state-of-the-art imaging filters. More importantly, the dark channel prior combined with the proposed filter possesses the highest potential for practical application due to its superior dehazing effect and time complexity.