Removal of rain in video based on motion and shape characteristics of raindrops

Rain degrades the performance of the outdoor computer vision system, which adversely impacts tracking, object detection and recognition. To solve this problem, a novel rain detection and removal method is proposed in this paper. Firstly, raindrops are considered as the motion component and extracted by FastICA algorithm. Secondly, raindrops are identified based on its properties of area, direction and width. Finally, a two-step removal method is proposed to replace pixels covered by rain with estimated background pixels. Experimental results show that the algorithm can effectively remove the raindrops in various videos. Compared with traditional rain removal methods, it does not require any parameters about rain, camera and scenes, and only needs three successive frames to recover a raindrops free video.