Focusing on the problem of moving object detection in moving camera, a new method for the estimation of global motion from compressed image sequences is proposed. With the global motion estimation as basis, utilizing the similarity of the background micro blocks, the algorithm can build background micro blocks sets speedily. Adopting the common four-parameter global motion estimation model, the motion parameters are estimated. By computing and filtering the motion vectors residual, the moving object is detected. The algorithm can utilize the motion information contained in the MPEG-4 video coded stream without decoding the coded stream completely, so it can improve the detection efficiency and effect dramatically.Testing results validate its advantages in global motion estimation.