Fast Video Object Segmentation in H.264 Compressed Domain

In this paper we proposed a real-time video object segmentation algorithm that works in the H.264 compressed domain.  The algorithm utilizes the motion information from the H.264 compressed bitstream to identify background motion model and moving objects.  In order to preserve spatial and temporal continuity of objects, Markov random field (MRF) is used to model the foreground field.  Quantized transform coefficients of the residual image are also used to improve segmentation result.  Experimental results show that the proposed algorithm can effectively extract moving objects from different kind of sequences.  The computation time of the segmentation process is merely about 15ms per frame for CIF size frame, allowing the algorithm to be applied on real-time applications.


Results in MPEG-1 video format




Mother & Daughter




Lab 1

Lab 2