Abstract: Because the traditional saliency model can not balance the fast motion scene and the slow motion scene，improves the movement feature extraction algorithm in modeling method of saliency model，promotes the accuracy of movement saliency subgraph. Considering the human visual system may concentrate more attention on partial region of the picture，applies the improved saliency model into the redundant slices coding for H.264 video，carries out redundant slices coding to the region paid attention by human eyes，so improves the error resilient performance of video stream transmission. The experimental results show that the proposed algorithm can remarkably improve both subjective and objective quality of the decoded picture in the packetloss environment.

Key words: redundant slices, error resilient coding, saliency model, visual attention, H.264 video