The Research Of The Moving Object Segmentation Technology Based On MPEG-4

Due to the development of streaming media and emerging of content-based functionalities in MPEG-4, moving object segmentation technology becomes a popular research in video field. The technology of moving object segmentation, significant part of multimedia application, not only have a widely application in these fields, such as video coding, retrieval, browsing, virtual reality and multimedia communication in the future, but also play an important role in computer vision, pattern recognition and so on. However, because of the complexity of the video's contents, no general method to video motional object segmentation is accepted now.Up to the present, both at home and abroad, the research of the moving object segmentation is mainly in the pixel domain or in the compressed domain. However, there are some defects in the pixel domain, such as the large amount of computation, the high complexity and the difficult to meet people's real-time nature. Through a large amount of research and learning, this paper is going to propose a method which using the new three-step algorithm form the motion estimation to extract the motion vectors form the target between prediction image. After using the accumulator and the filter to deal with the motion vectors, we make use of the SOFM neural network to cluster these motion vectors, so finally we extract the moving object from the target image.The method this paper proposed not only decrease the computations, but also improve the need of the people's real-time nature. As the SOFM network is an adaptive network, we do not need to know the number of the moving objects comparing with the other network, it can finally decide the number of the moving objects, so the SOFM network will be widely used in the field of multi-moving-objects segmentation.