Effective moving object detection in H.264/AVC compressed domain for video surveillance

In this paper a novel approach is presented to detect moving object in H.264/AVC compressed domain for video surveillance applications. The proposed algorithm utilizes the information from the H.264 compressed bit stream to reduce the computational complexity and memory requirements. In order to exploit the spatial and temporal consistency of moving object, a Markov Random Field (MRF) model is employed to detect and segment moving object based on motion vectors and quantization parameters (QP). The size of the blocks (in bits) are also used to improve the detection result. Experiments show good performance achieved by the algorithm, and the moving object can be detected effectively from the compressed video sequence.