Video's global motion (camera motion) behaves time-relativity ofvideo sequence. It have more higher semantic information than any other video's character. In order to realize the video segmentation quickly,How to acquire video's global motion more efficient is very emergency now, a novel method for video's global motion estimation based on a three-parameter model,the new model uses less parameter to describe and estimate global motiom,its estimation results are still accurate adequately. An approach for block judgment is presented before motion field calculation ,gradient information is analyzed to judge the blocks, we should make the global motion estimation and compensation first , and then segment objects in terms of the static one. Extensive experiments show the effectiveness of this technique. |