Abstract: Video object segmentation is the most important technology in the virtual advertisement system.For increased accuracy and efficiency，this paper proposes the use of loopy belief propagation.First，background，shadow and foreground are assumed to be modeled statistically.Then，the neighborhood dependencies are modeled with a Markov Random Field（MRF）.Finally，based on belief propagation，the video objects are segmented effectively.Experiment results show that the algorithm has the good performance in the virtual advertisement system.