Research & The System Implementation Of Video Compositing Algorithm

With the development of science and technology as well as the improvement of people's living standards, we can easily obtain the different effects of video or image materials. The video synthesis technology is adopted, in order to proceed effective post-processing to these materials, and generate to meet the specific requirements of a video or image. At present, the video synthesis technology is researched mainly based on general image synthesis algorithm. Most of these algorithms are focusing on the seamless synthesis and ignore the color retention of the source image. On the other hand, these algorithms are large count in terms of cost for video. Therefore, puting forward a video image synthesis algorithm that owns characters of low overhead, seamless synthesis, and color appropriate to retained has important research significance.This paper based on the research and the system implementation of video compositing algorithm as the research subject for fixed background video, mainly researchs the preprocessing of video compositing, color adjustment for image composition algorithm and develops a video compositing software system. The main research contents are as follows:Firstly, this paper researchs the preprocessing of video compositing. The purpose of preprocessing of video compositing is to obtain the accurate foreground mask of the video. This paper uses the Visual Background extractor(Vi Be) algorithm for the preprocessing of video compositing. Since the original Vi Be algorithm can't be accurate segmentation the local motion objects, produces debris and is sensitive to illumination changes, this paper puts forward an improved Vi Be algorithm. This paper improves the segmentation mask and update mask of the original Vi Be respectively, and achieve the purpose of filtering segmentation fragments. By inhibiting the spatial propagation of background seeds in the foreground objects, the foreground which is slowly changing or local static segmentation has greatly improved. The influence of illumination change is reduced by the image gray scale normalization method.Secondly, this paper improves the Poisson image editing technique. Poisson image editing technique suffers from two main problems: bleeding artifacts and color bleeding, because of the sole dependency on the intensity of the boundary pixels in the target image. In this paper, the proposed modified Poisson blending(MPB) algorithm reduces the problem of the bleeding artifacts by ensuring a more balanced dependency on the boundary pixels in the target and the source images. The problem of color bleeding is reduced by using alpha compositing between two images generated by the MPB process.Thirdly, to solve the problem of the current image synthesis algorithms are large count in terms of cost, the source image's color can't be retained and the composited images are not natural, this paper puts forward the color adjustment for image composition based on convolution pyramids(CAICCP) technique. O ur technique can realize rapid images composited using convolution pyramid fast approximate estimate the Laplace membrane, and according to some automatically-computed interpolation parameter based on perceptual color differences between the interested area of the source image and the target image, the user can also manually set parameters to decide the degree of color preservation. According to experimental comparison, the effect of the composite image generated by our technique has greatly improved compared to currently available methods.Finally, using the algorithm researched in this paper we develop a video compositing software system. The system can achieve the function of general video images seamless composition, and generate a new video which has new special effect contents. The experiment results show that the effect of composited video produced by improved Vi Be algorithm and the CAICCP algorithm is improved greatly compared to currently available technique, and the video compositing software system is strong practicability.