A motion detection scheme for video surveillance

A scheme for motion detection in compressed code stream for video surveillance is presented in this paper. The scheme reads out signs of the discrete cosine transform (DCT) coefficients from the compressed video code stream and detects motion by counting the number of DCT coefficients that change their signs in a pair of DCT blocks corresponding to consecutive frames. A probability model and nonlinear filtering is used for eliminating the disturbance of noise in the background and luminance change, respectively. Experiments on various types of video sequences show the proposed scheme can perform motion detection correctly with low computational cost.