A New Pedestrian Detection System using MPEG-2 Compressed Domain Information

M. Coimbra and M. Davies (UK)


Computer Vision, MPEG-2, Compressed domaininformation, Surveillance


Surveillance is one of the applications of the video processing field where algorithmic speed is vital. Our proposal exploits the fact that popular video compression standards such as MPEG-2 have large amounts of compressed domain information that can be accessed with minimal decoding. A very fast algorithm is obtained that uses motion vectors and DCT coefficients to effectively detect moving pedestrians using real videos from underground train CCTV surveillance systems.

Important Links:

Go Back