Removing rain and snow in a single image using guided filter

Dynamic weather conditions, such as rain and snow, degrade the performance of vision algorithms, which is served for video surveillance and analysis tasks. Many methods have been proposed to reduce the effects of raindrops and snow flakes that are regarded as noises in each frame of a certain video. However, the removing of rain-drop or snow-flake is quite challenging since each of them affects only on a very small region of an image, thus, leading us a rub to determine which region should be smoothed and which should not. In this paper, a guided filter, by which no pixel-based statistical information for detecting rain and snow are needed, is introduced in the removal stage. The results show that it may achieve better performance. In addition, the degradation-restoring algorithm is suitable for both single images and videos.