Video satellites have great potential for military and civil use due to their capabilities in the dynamic remote surveillance. However, the video data collection rate is far higher than the bandwidth between satellite and earth. Video coding is the technology to compression the video data to a low bitrate for transmission, by eliminating the local spatial/temporal redundancies, but there is still a gap between the bitrate of the video data and the satellite-earth bandwidth. Long-term background redundancy is a new type of redundancy existing in the remote satellite video data, caused by the repeated recording of the same place. This type of redundancy becomes significant as the video data of the same place increases. In this paper, we first discuss the factors causing the change of the background, and propose the long-term background reference library. After that, we propose the method for the background reference generation and the coding framework for satellite video data. Experimental results show that 64.38% reduction on video data bitrate can be achieved by using the proposed method, compared to the H.264 video coding standard. The proposed method will boost the applications of video satellite data in the surveillance field.