Motion Object Detection Based On MPEG-4 Video

Moving object detection is one of active branches of imaging processing and computer vision, which studies how to extract moving object region and estimates the concerned moving parameter in serial video image. This technology has been applied widely in military application and video surveillance. This paper does some research on moving object detection as follows:Focusing on the problem of moving object detection on MPEG-4 compressed streams captured by a stationary camera, an algorithm is presented to detect the moving object using motion vectors and DC coefficients of the motion information in MPEG. The algorithm is to firstly mark the moving regions based on the motion vectors, which has been filtered by using cascade filter, then implement a local object detecting by using DC coefficients of each marked block and calculating the difference between the central block and its neighbors'. Finally, the border is rectified by extending the objects'region and comparing border blocks'DC coefficients with the average background's value of the new region. Testing results show that minimal decoding on compressed bit stream is performed and the detecting effect can be dramatically improved.As moving object detection under moving camera is concerned,GME(Global Motion Estimation)is an important method, which is widely used in moving object detection. A new method is proposed for the estimation of global motion from compressed image sequences. As known to all, GME is based on proper construction of background image. From this point, the algorithm is implemented by searching and constructing a background image efficiently. With regard to global motion models, four-parameter estimation model is adopted. Finally, according to the estimated error, the moving object is extracted. Using this method, it is not necessary to fully decompress MPEG video sequence. Both the efficiency and detecting effect is improved dramatically.