Rain removal from video using dark channel prior and guided filter

Raindrops in images or videos degrade the visibility and lead to a bad output result for most digital imaging techniques. In this letter a novel rain removal method is described. Firstly, a matting model is proposed based on the imaging model of raindrops. The parameters in the matting model are analyzed such that we can solve the matting equation by use of dark channel prior. The matting model could remove most of blurred raindrops. Secondly, due to the spatial property of raindrops, we can use guided filter to remove those clear linear rain streaks, which are very close to the camera. In our method, the removal procedure does not need lots of frames. Also the alignment of input images is not necessary. The result shows that our algorithm has good performance in most rain-affected videos especially for those including dynamic scenes.