This paper proposes a low-complexity technique to structure video sequences and generate corresponding meta data supporting adaptive presentations for arbitrary communication terminals. The algorithm structures a video sequence into segments of coherent camera motion. The structured video is presented by representative key-frames, which are characterized by background and moving or stationary objects. The realized prototype incorporates algorithms for camera motion estimation, spatial-temporal segmentation, object tracking, key-frame selection, and meta data (MPEG-7) generation.