Moving object detection in the H.264 compressed domain

A new method for the automatic detection of moving objects directly in the H.264 compressed domain is proposed in this paper. The method takes advantage of the multiple block sizes, which are used during the inter-mode decision by the H.264 encoder. The different blocks/sub-blocks are combined with their associated motion vectors in order to denote a moving object. The method works in the compressed domain as the block-sizes and the motion vectors can be found by partially (entropy only) decoding the H.264 bitsream. Finally, the method can optionally extract the details of the moving object by fully decoding only the macroblocks that the object consists of.