Segmentation Algorithm Based On The Panorama Of Moving Objects

Video extraction is a very hot research area in both computer graphics and computer vision. The goal of video extraction is to separate the moving object from the background. With the technology of video synthesis, the algorithm can merge the extracted foreground with another scene to make a new video. With the fast development of computer graphics and multimedia technology, video extraction is widely used in movie production, military affairs and so on.The traditional extraction algorithms which are based on the hypothesis that the camera is fixed mainly depend on the image information (such as color, edge, texture and so on), models and prior knowledge to separate foreground from background. In this paper, we propose a novel video extraction algorithm based on panorama. The algorithm not only can handle camera panning situation but also cope with camera movement with zoom. Ego-motion of the camera is compensated by a robust technique and build panorama of background using the motion parameters between each frame. Extraction of moving objects is carried out using space coherence and polling technology. The automatic computation of the panorama and the extraction of foreground object are compared with the current literature and illustrated with real sequences experiments.Furthermore, we find that the extraction algorithm is also applicable in video synthesis to make a new video.