The Video Matting Based on Background Reconstruction and Prediction
Yang Shen, Lizhuang Ma · International Workshop on Education Technology and Computer Science · 2011
In this paper, we propose a new video matting method based on background reconstruction and prediction. Different from image matting technique, video matting can benefit from temporal consistency. so we can predict or reconstruct the background from other frames. By known background, we can get accurate trimap and high quality alpha matte. We detect the movement of background objects, so we can reconstruct and predict the occluded background object behind the foreground in each frame. After background reconstruction and prediction, a simplified robust matting algorithm is used to extract the alpha matte. Experiment results show that our method achieves ideal results in some scenes where old methods get poor results.