Research And Implementation Of Video Matting And Video Stylized Rendering

With the large popularity of digital products, the rapid development of communication technology and the considerable progress of entertainment and multimedia technology, the processing of video files has become an industry hot spots. The video stylized rendering technique which is widely applied, it is an important method of showing art by using computer technology.Matting technique have an important role in the field of image processing and television production, etc. It is a technology of extracting the foreground objects which have arbitrary shape from the source image out. Video matting is an extension of the traditional natural image matting, it can divided the video into a stream of foreground, a stream of background and a stream of mask, and it has a high accuracy. Video is continuous and it’s a combination of temporal and spatial correlation with a series of images, thus keep the temporal consistency and reduce human interaction is the key of the video matting technology, Video matting algorithm which used in this paper is based on Video SnapCut algorithm and make improvement to ensure the efficiency of the algorithm which can get a better results.Image,(video) stylized rendering is belong to non-photorealistic rendering (NPR) technology, which goal is to simulate the artistic effect by computer. Keep the temporal and spatial consistency is a challenge in the video stylized rendering, Video Painting algorithm is based on motion layer operation, it’s a better algorithm than make the method of stylized images directly applied to the resulting video frame lead to the temporal discontinuity problem. In this paper, the improved video matting algorithm is applied to the motion layer extracted in the Video Painting algorithm, thereby obtaining a better layer segmentation and improving the robustness of the algorithm.Firstly, paper introduces the developing situation of the video matting technology, make the description and analysis of the existing video matting algorithm, then detailed overview the Video SnapCut matting system. In this paper, we base on the thought and framework of the algorithm, and improve on many ways of the algorithm, such as we used the watershed algorithm replacement the Lazy Snapping algorithm which used in obtain the initial foreground edge in the key frames, thereby reducing the human interaction and improve the robustness of the algorithm, we can get a more accurate initial edge; we used the SURF algorithm replacement the SIFT algorithm for detecting the feature points in a video frame and so on, thus the overall efficiency of this algorithm is higher.Secondly, the paper introduced the video stylized rendering technique, and detailed analysis the motion-based layer Video Painting algorithm. When extract the motion layer which is the most important part of the algorithm, we used the improved video matting algorithm in it, as bring in the local classifier which can improve the robustness of stylized rendering system greatly, namely it can also have a good effect when faced the complex background video.Finally, this paper implement a video stylized rendering system which has high robustness. This system combines the NPR technology and video matting technology, it can obtain the accurate motion layer segmentation through the video matting, then use the NPR technique on the different motion layers for rendering, we can get a video with artistic style at last.