Fast Haze Removal Algorithm for Surveillance Video

In this paper, we present a new approach to remove haze from surveillance video sequences. The approach first extracts the background image through frames differential method, and then estimates the atmospheric light and transmission map from the background image, finally renders haze-free video according to the haze image model. The main advantage of the proposed approach is its speed since this approach adopts a ‘universal strategy’ that is applying the same atmospheric light and a single pseudo-transmission map to a series of video frames. Experiments and performance analysis demonstrate that a good haze-free video can be produced effectively and efficiently.