Efficient Global Motion Estimation Method for MPEG Video

Motion estimation plays an important part in the analysis of video sequences for the purposes of information extraction.In order to process video data effectively,camera motion extraction from compressed video is often required.In this paper,an effective global motion estimation algorithm based on the moving background derived from MPEG compressed data is presented.The algorithm uses the predictive error DC-image to extract the moving background of the current frame,and then exploits a new global motion estimation method based on a 6-parameters motion model to calculate the camera parameters.In the evaluation experiments,five video test sequences with global motions and three video sequences without global motions are used to test the proposed algorithm,and compared with an existing method.The experimental results show our algorithm can estimate global motion parameters more effectively and exhibit more robustness than the existing algorithm.